Development of advanced high heat flux and plasma-facing materials

Re is by far the most important alloying element for W with respect to ductilization. The maximum solubility of Re in W is 37 at% Re at 3000 °C; this decreases with decreasing temperature, reaching 27.5 at% at 1600 °C [48]. W–Re alloys have gained outstanding importance since it was shown that they exhibit a significantly lower DBTT [49], and are even stronger than pure W at high temperatures [50]. Seeing that the main shortcoming of W is its low-temperature brittleness, it is not a surprise that the interest in alloying by Re has considerably increased. Furthermore, Re additions have a beneficial impact on the reduction of the degree of recrystallization embrittlement [51], significantly improve weldability, and when compared to pure W, W–Re alloys show superior corrosion behavior [22].

One of the first investigations of the supreme properties of these alloys was conducted in the mid 1950s by Geach and Hughes [52]. They reported that W–Re alloys show a good temperature strength (that originates from Re) and a very high ductility, that can be related neither to the pure W nor to the Re. The room temperature (RT) ductility of W alloys varies with rhenium content, reaching a maximum at about 30 wt%. They attributed the improvement in low-temperature ductility to the mechanical twinning which allows plastic deformation at low stresses, thus avoiding the increased locking of dislocations at low temperature and high strain rates [49].

A detailed fracture mechanical investigation was performed by Mutoh et al [53], focusing on the fracture behavior of W and its Re alloys (W–5 wt%Re and W–10 wt%Re) at elevated temperatures. The results show that for all three materials K_Ic is about the same at RT with the values of 12–14 MPa m^1/2 [53]. With increasing temperature, Re alloys tend to be tougher up to 1000 °C, when a decrease is observed. Corresponding to the results of the fracture toughness values at RT, fractographic analysis of all the materials reveals a quasi-cleavage brittle fracture. At 800 °C, a ductile dimple fracture is visible on the fracture surfaces of the Re alloys, indicating that the DBTT has already been exceeded. In the case of pure W, the fracture surface still has a dominant quasi-cleavage fracture with no dimples. At very high temperatures, pure W experiences an embrittlement of the intergranular regions and crystal growth due to recrystallization. However, the materials containing rhenium show no significant recrystallization and embrittlement of the grain boundaries, leading to a mild reduction of toughness values. Based on the obtained results, adding Re to W has a beneficial impact on the improvement of the high temperature toughness, as well as on the decrease of DBTT.

In the work of Gludovatz et al [54], the effects of the microstructure and the production route on the fracture toughness of W based materials were investigated. For all materials, as expected, an increase of fracture toughness was observed when raising the temperature. However, W–Re alloys show significantly higher levels of K_Ic, and more pronounced plastic deformation, in comparison with other materials. Even at room temperature a very distinct plasticity is revealed in electron back scatter diffraction (EBSD) scans, along the crack path of the propagated crack. At higher temperatures, plasticity appears in the wider vicinity of the crack flanks (figure 1).

Zoom In Zoom Out Reset image size

The main focus in [55] was the investigation of the fracture behavior of W–26Re alloys between RT and 900 °C. The experimental results reveal a high anisotropy of fracture toughness in the case of the as-forged material, where the conditional fracture toughness was changed by a factor of almost two, when changing the crack plane and crack propagation direction with respect to the microstructure. Recrystallized samples are less tough compared to specimens made of rolled material; nevertheless, the fracture toughness values of recrystallized W–Re are well above those of recrystallized pure W. The kind of recrystallization that might happen in future fusion reactors completely changes the microstructure, affects the good mechanical properties, and lowers the fracture toughness. However, in [55] it was shown that even though the fracture toughness is lowered by almost 50% at room temperature and 300 °C, after recrystallization, this material is still superior over pure, recrystallized tungsten, by a factor of three.

So how can this advantageous influence of rhenium alloying to tungsten be explained? What is the mechanism(s) that enhances ductility and toughness, and what kind of physical processes lie beneath this effect?

A decade after the discovery of improved properties for W–Re, Booth et al published a paper reviewing several possible factors that could explain this positive influence [56]. One fact is the promotion of twinning in Re-based alloys with W and Mo, which was identified as a mechanism aiding deformation, e.g. by reducing stress concentrations. However, this cannot be the sole factor making these alloys tough.

Segregation of Re to grain boundaries was also discussed in [56]. At that time, the conclusions were rather drawn indirectly from a change in fracture morphology, going from cleavage fracture for pure materials, to grain boundary fracture for Re alloys. However, two years later Gilbert [57] found a change from grain boundary fracture in unalloyed W to cleavage fracture for Re alloys. From the authors' experience [58] it can be said that it is difficult to link the ratio of inter- to transcrystalline fracture with alloying (or impurity) content. It is influenced to a greater extent by microstructural features such as grain size and grain shape.

Besides the improved properties, many authors mention the solid solution softening phenomenon which can be observed at low temperatures [27, 59, 60]. When adding small amounts of rhenium, there is a significant decrease in hardness. However, this effect diminishes on increasing the Re content and the temperature; it completely disappears above a homologous temperature T_m of 0.16 [27]. The phenomenon of solid solution softening is not peculiar to Re alloys—it is a common feature of bcc materials. Keeping in mind that W–Re alloys also show markedly good mechanical properties at high Re contents and elevated temperatures, there have to be (at least) two mechanisms working that ductilize and toughen these alloys. For applications, the solid solution softening is of less interest, as this effect vanishes for higher temperatures. A thorough review of this effect is given by Pink and Arsenault [59].

What seems to promote these improved mechanical properties of W–Re alloys is assigned to the change in the dislocation properties, in particular screw dislocations of ${{}^{1}}{{/}_{2}}\langle 1\,1\,1\rangle$ type. Gröger et al [61] performed computer simulations based on bond order potentials, and showed that $\left\{1\,1\,0\right\}$ glide planes are the preferred glide planes for pure W and Mo at low temperatures. An explanation of the Re ductilizing effect was then given by Romaner et al [62] by means of the density functional theory (DFT) calculations of ${{}^{1}}{{/}_{2}}\langle 1\,1\,1\rangle$ screw dislocation in W–Re alloys. It was demonstrated that alloying with Re leads to crucial changes in the interatomic bonding, thereby affecting the screw dislocation core structure. Re additions lead to a pronounced asymmetry in the $\langle 1\,1\,2\rangle$ direction. The existence of the transition from a symmetric to an asymmetric core is revealed; this causes a change in the preferred slip plane [62], and has a significant influence on the glide behavior of the dislocation. Whereas the glide in pure W is uniform on one glide plane, the glide plane changes for W–Re alloys, resulting in an overall movement on $\left\{1\,1\,2\right\}$ planes. Additions of Re affect not only the slip plane, but also the Peierls stress ${{\sigma}_{\text{p}}}$, which is lowered. ${{\sigma}_{\text{p}}}$ is the shear stress required to make a dislocation glide in an otherwise perfect crystal without the assistance of thermal activation. In agreement with this are the results of many experiments where the plastic behavior and the glide planes of W and various W alloys were investigated [63–65]. At room temperature for pure W, in both $\left\{1\,1\,0\right\}$ and $\left\{1\,1\,2\right\}$ slip systems, a slip has been observed. Adding Re promotes slip at room temperature on $\left\{1\,1\,2\right\}$ planes. The publication by Li et al demonstrates and explains this increase in number of slip planes, combining both experimental work and computer simulations [66]. Similar simulations were performed for alloys of W with Ta and V, but no change in the screw dislocation core structure was found, which is in agreement with the experimental results collected over the years.

To summarize, the ductilizing effect can be attributed to the change in core symmetry, the decrease of ${{\sigma}_{\text{p}}}$, and the increase of the number of slip planes. Hence, the origin of this phenomenon is intrinsic, and results from the filling of the d band, which modifies interatomic bonding, leading to enhanced dislocation mobility and improved plastic deformation.

It becomes clear that ductility and toughness enhancement of W by formation of solid solution is possible by the addition of Re. Many experimental results show e.g. improved toughness and lower DBTT, and computational results give the explanation. However, despite the strong technological relevance of the Re ductilizing effect its use is limited by the fact that Re is a rare metal. For fusion energy application, W alloying with Re is inhibited due to the element's low availability. Nevertheless, knowledge on the effect of Re alloying is important for fusion applications. W atoms partially transmute into Re due to neutron irradiation, and therefore materials' properties might change with increasing neutron dosage.

Iridium is rarely mentioned in the literature as a typical W alloying element. Nevertheless, its electronic configuration, the neighbourhood to Re in the periodic system of elements and the W–Ir phase diagram indicate some promising effects on W. The solubility of Ir in W decreases with decreasing temperature—the maximum of about 10 at% is at the peritectic temperature (2540 °C) and decreases to about 4 at% at 1810 °C [48].

Luo et al investigated the mechanical properties of W–Ir alloys in two publications in the early 1990s [67, 68]. In the first paper, the scope was the investigation of a solid solution softening mechanism of Ir in W at room temperature [67]. It was deduced that even with small amounts of Ir the properties change noticeably. The concentration of 0.4% Ir corresponds to the maximum in solid solution softening, minimum strength and maximum ductility. In comparison to the alloy containing Re, these effects are more pronounced, and the iridium is shown to be more efficient in terms of solid solution softening.

In the second paper from Luo et al, the investigation of the effects of Ir concentration and the test temperature on the strength and fracture behavior of W–Ir alloys at ultrahigh temperatures was performed [68]. Iridium shows a moderate strengthening effect in the temperature range 2200–2600 K, but with the increase in temperature the effect decreases and finally vanishes at 2600 K. The increase in tensile properties of the alloy is proportional to the concentration of added Ir (see figure 2 in [68]).

As the experimental data for tungsten–iridium systems are limited and very little is known, in order to make final conclusions regarding the beneficial effect of Ir addition, further investigations of mechanical properties are necessary. However, from a technical point of view viz. fusion applications, iridium is assumed not to be significant due to its very low abundance. Yet if about 0.2 at% would be sufficient to obtain a substantial improvement in mechanical properties, then alloying by iridium might be reconsidered for applications in fusion environments.

The W–V system has a continuous series of solid solutions with a bcc structure, occuring between pure V and pure W at all compositions [48]. They exhibit high strength, good fabricability and good corrosion resistance. Experimental results of mechanical properties and the fracture behavior of W–V alloys are scarce. Based on the data from the Charpy impact testing at different temperatures, it seems that alloying W with V is not a promising avenue of toughness and ductility enhancement. No improvement in the DBTT can be observed (figure 2), and for temperatures below 1000 °C when compared to pure tungsten, all depicted alloys are more brittle [21].

Zoom In Zoom Out Reset image size

Additionally, it should be noted that a density functional theory (DFT) investigation on V alloying was conducted, but did not reveal any significant change in the dislocation core symmetry [69]. Hence, using V as alloying element will not be an option for the ductilization of W.

However, the idea of using V as one of the compounds is not completely discarded. In [70] W–V composites have been explored for a possible medium temperature application for fusion technology. The proposed materials show promising properties, and results of the numerical analysis reveal a distinct potential of the material—e.g. by varying the content of V, the thermal expansion coefficient can be changed over a large range.

Furthermore, the fracture behavior of the HPT-deformed ultra-fine grained (UFG) W–V material was investigated by Wurster et al [73]. The results show that the fracture behavior is not markedly different to UFG pure tungsten, with a strong anisotropy regarding testing direction [71]. The material homogenized at high temperatures shows inferior fracture behavior. With these new results available it was stated that W cannot be ductilized and toughened by alloying with V [20]. Composite production, for medium temperature application, seems to be the only viable option.

The tantalum–tungsten system is characterized by complete miscibility in both solid and liquid states over the entire range [48]. Tantalum alloys, with a few percent of tungsten, have gained great technical significance—especially for structural materials in corrosive media. Advantageous properties are due to the combination of high elasticity and good corrosion resistance of Ta, and solid solution strengthening of W [22].

The possibility of using tantalum as a tungsten ductilizer was also explored. Referring to figure 2, the results of Charpy impact testing are not promising—the increase in the amount of Ta leads to a decrease in the fracture toughness of W–Ta alloy. The DBTT in Charpy tests seems to be even higher.

Wurster et al [73] performed a detailed investigation of industrially produced solid solution W–Ta alloys. An increase of fracture toughness was observed with increasing temperature, as well as a decrease with increasing Ta content. Strong dependence of fracture mode on Ta content is evident since, for W–10Ta, the maximum tested temperature was not sufficient for transition from transcrystalline to intercrystalline fracture to take place. For this special testing geometry, this is a sign of an increasing grain interior embrittlement with increasing Ta content. When compared to pure tungsten, transcrystalline fracture toughness of W–Ta is significantly lower at high temperatures.

Based on these results and several other publications [20, 21], it is clear that tantalum additions lead to a decrease in fracture toughness and workability. Therefore, these alloys will not be proposed for fusion applications.

Based on experimental and computational results, it can be concluded that the ductility of tungsten can be improved by formation of solid solution when adding Re, Ir and—continuing the sequence—maybe Os. In contrast to the thorough research on the Re effect, alloying by Ir and Os has not been investigated in detail, therefore the true extent of possible improvement of mechanical properties and minimal needed alloying content is unknown. For a large, technical scale application like fusion reactors, these elements seem to be ruled out mainly due to their low abundance and their high price. Therefore, when treating the problem of room temperature brittleness of tungsten, the solid solution approach is not an option for fusion application, and other solutions need to be found.

WL10 is a nanostructured tungsten material with 1 wt% La₂O₃ particles dispersed in its metal matrix. The microstructure of the material, its texture and grain size strongly depend on the manufacturing history. During the production steps (e.g. rolling) the lanthanum oxide particles, which initially have spherical shape, become elongated along the rolling direction, forming needle-like structures. In a typical industrial product, rod material with a diameter of 10 mm and a final hot forging step (degree of deformation  >80%), the size of the lanthanum oxide needles is about 400 nm in diameter and 20 μm in length [74].

In the work done by Rieth and Dafferner [74], Charpy impact tests on commercially available W and WL10 rod materials were performed. In order to achieve optimum Charpy properties, due to the anisotropic microstructure, testing specimens were fabricated with the long side along the rod axis and with the notch perpendicular to it. The experiments were performed up to 1100 °C, and DBTT was determined to be $800\pm 50$ °C for pure W, and estimated to be $950\pm 50$ °C for WL10. The fact that the already insufficient fracture characteristic of pure W is further reduced by adding La₂O₃ is even less promising. Fully ductile behavior was only observed for tungsten at 1050 °C, while all the other specimens fractured in a brittle manner. Thus, the value of a DBTT for WL10 can only be estimated. WL10 rods behave more like uniaxial fibre-reinforced materials, where the crack propagates along the needle-like La₂O₃, which yields a further rise of the DBTT compared to pure W. An additional, detailed investigation of Charpy properties was performed on pure and WL10 tungsten rod materials in order to examine the influence of microstructural characteristics like grain size, texture and anisotropy, as well as notch fabrication [75]. All of these materials exhibit brittle behavior at temperatures below 600 °C, where the weakest links in the microstructure are the grain boundaries, leading to intergranular fracture. The authors conclude that, regardless of the beneficial impact of La₂O₃ on W materials (such as an improvement of the processability and tensile properties, the suppression of recrystallization and a slight strengthening in creep [75, 76]), fracture characteristics compared to pure tungsten are not enhanced.

However, an appropriate optimization of the production method leading to a homogeneous grain distribution and a reduction in size of dispersed particles could improve ductility to some extent. Yan et al [77] performed a systematic investigation of the effect of working processes on the microstructure, fracture behavior and DBTT, comparing pure W and WL10. The results obtained in [77] distinctly differ from those in [76], since it was found that WL10 has a lower DBTT than pure W. Impact toughness of pure tungsten can be improved by an increase in relative density, since pores could act as crack initiators. Density measurements that were also performed indicate that swagging before rolling could lead to higher fracture strength by eliminating residual pores from the matrix. According to their opinion, swagging  +  rolling as a production method could be successful, in combination with lanthanum oxides additions, and could lead to enhanced mechanical properties of tungsten materials.

When discussing nanostructured and ultra-fine grained materials, it is very important to emphasize the stability of the microstructure at elevated temperatures, since beneficial mechanical properties would deteriorate if recrystallization should occur. Potassium-doped tungsten (WVM, W vacuum metallizing) is a nanostructured tungsten material doped with potassium (0.005 wt%). The potassium bubbles are known to retard recrystallization [78], and therefore this is a very promising option for nuclear fusion applications. The effect of doping tungsten with potassium is well known from the wire industry [79]. During processing, deforming and annealing, strings of fine, nanometer sized potassium bubbles are formed and located in grains as well as on grain boundaries, effectively impeding dislocation and grain boundary movement [80].

In [75], the DBTT for WVM rod material was determined to be around 1000 °C, which is even higher than the value for pure tungsten. For the materials used, it was also determined that rod microstructure is more favorable compared to rolled plates, indicating that underlying microstructure has a significant influence. Gludovatz et al [54] investigated the fracture toughness of various tungsten materials (including WVM) by the means of three-point bending, double cantilever beam and compact tension specimens. As expected, all the materials showed an increase of fracture toughness with increasing temperature, and WVM even has higher values of K_q than pure tungsten, at tested temperatures of 400 and 800 °C. Faleschini et al investigated the toughness of tungsten materials, taking into account the influence of cold work [71]. Strongly depending on the degree of deformation, the fracture toughness was found to be increased for L-R orientation (according to ASTM E399) compared to sintered materials over the whole temperature range. For C-R orientation the toughness was comparable at low and worse at higher temperatures than for sintered materials.

So, it is clear that toughness could be improved to some extent by dispersion-strengthened tungsten by appropriate manufacturing processes. A high degree of deformation, as well as an alignment of grains along a certain contour, has a considerable impact on ductility and fracture toughness of tungsten materials. This indicates that materials' microstructural characteristics like grain size, anisotropy and texture play an important role in overall mechanical properties. During processing, dispersoids like La₂O₃ or potassium bubbles affect the evolution of the microstructure and significantly increase the stabilization of fine microstructures with improved properties.

The research on CuCrZr alloys is well covered in the publications [90–93], and we would like to direct the reader there. Further details on Cu-based heat sink materials are discussed in a separate manuscript within this special issue ¹³ . Here, we briefly summarize the conclusions from those original publications:

• Substantial thermal recovery of irradiation damage, if the temperature of irradiation (1–10 dpa) is higher than ca. 250 °C. Above 300 °C; ductility is fully recovered.
• Strong hardening and embrittlement when irradiated below 150 °C. The uniform elongation is negligible.
• In the transition temperature range between 150 and 250 °C, considerable capability of total elongation (ca. 6%) before rupture. However, the load bearing capability decreases with strain.
• Significant loss of strength at elevated temperatures above ca. 350 °C; owing to over-ageing and irradiation creep after long-term thermal exposure.
• Sufficient fracture toughness even after irradiation. Toughness decreases with temperature, exhibiting only slight dose-dependence.

The research on tungsten divertor materials is reviewed in a publication by Rieth et al, and we would like to direct the reader to reference [74]. The main conclusions can be summarized as follows:

• Ductile-to-brittle transition (DBT) behavior occurring between ca. 600 and 900 °C. The DBT temperature is sensitively dependent on strain rate, microstructure and irradiation dose.
• Marginal capability of plastic deformation below the DBT even without irradiation.
• Recrystallization at ca. 1300 °C; leading to significant embrittlement and softening.

The features of the mechanical performance of CuCrZr and W described above suggest that both materials will have to be used within a respective allowed operation temperature range, to avoid brittle fracture or plastic failure. Generally, this requirement is strictly respected in design practice, particularly for structural materials. In the present case, the heat sink tube is the structural part. Thus, the desired temperature window for CuCrZr alloy should be considered as mandatory. In terms of ductility and strength, the optimal temperature range lies between 250 °C; and 300 °C.

However, the paramount requirement of power exhaust is to avoid any coolant burnout accident, even under slow transient events, which sets the upper bound for the coolant temperature. In the case of current water-cooled divertor design for DEMO, this upper limit is roughly 150 °C. Moreover, the corrosion-erosion issue requires that the coolant temperature should be kept as low as possible, say, below 150 °C; [94]. Therefore, the structural design of the heat sink tube needs to accept a non-ductile operation regime below 250 °C; possibly down to 150 °C.

Regarding the W armor, there is no established design rule yet in the currently available design codes. As the water-cooled W armor is to be operated with a wide temperature range from far below the DBT temperature to far above recrystallization temperature, one cannot apply a desired temperature window to the monoblock type armor.

Figure 4 shows the axial (a) and the hoop component (b) of the thermal stress produced in the cooling tube after 10 HHF cycles at 18 MW m⁻²; (coolant: 150 °C) [95]. It is noted that thermal stresses (including the residual stress) are produced by the differential thermal strain mismatch between the W armor block and the CuCrZr cooling tube. Since the thermal stresses outweigh the static membrane stress due to the coolant pressure, we consider only the secondary stresses for our discussion about the structural reliability of the divertor heat sink. It is found that critical tensile stresses build up during the cooling phase, whereas compressive longitudinal stress prevails during HHF heating. The stresses are concentrated mainly in the upper region of the tube, whereas in the rest of the tube the stresses remain rather moderate or weak during both the HHF loading and cooling—see figure 4. The stress concentration appearing at the edge of the bond interface may possibly initiate fatigue failure or interfacial debonding. At the upper interface edge the stress state varies periodically from compressive to tensile upon the shut-off of the cyclic HHF loads (and vice versa upon the onset of the HHF loads). This means that the interface edge region will experience reversed loading that may possibly cause fatigue damage during pulsed HHF operation. In the bulk region, away from the edge, the stress intensity is relatively lower than in the edge region, but still high enough to trigger non-ductile failures of the CuCrZr tube being embrittled under neutron irradiation.

Zoom In Zoom Out Reset image size

A rigorous cyclic plasticity simulation shows that the plastic straining of the cooling tube occurs only during the fabrication process, provided that the CuCrZr alloy used for the tube does not undergo overaging (ripening of the precipitates) or irradiation creep (dissolution of the precipitates) during subsequent thermal exposure under the HHF loads and neutron irradiation, so that the initial prime-aged microstructure and thus the initial yield strength are preserved. In this case, considerable residual stress can be produced, due to the differential thermal strain mismatch during cooling in the joining process. Concomitantly, overall plastic yield occurs in the tube. The intensity of the residual stress depends on the thermal history—in particular, the effective stress free temperature of the fabrication process.

The presence of strong residual stress can significantly influence the further evolution of thermal stress during the HHF loading, as the secondary stress produced by the HHF loads is superposed onto the residual stress. Assuming a precipitation-hardened state of the CuCrZr tube, the high yield stress of the CuCrZr alloy leads to an accordingly high residual stress (ca. 300–400 MPa), which dominates over the secondary stress of opposite sign.

This feature is well demonstrated in figure 5, where the temporal variation of three principal plastic strains (expressed in a cylindrical coordinate system adapted to the tube geometry) is plotted over the first ten HHF load cycles at 18 MW m⁻²; (coolant temperature: 150 °C). It is seen that the stress state in the cooling tube already readily enters into the fully elastic shakedown regime upon the onset of the 'first' HHF load, so that no further plastic deformation takes place during the subsequent cyclic HHF loading after the first load cycle [95]. In this circumstance, the risk of plastic fatigue or ratchetting can be excluded. It is reminded that this condition can be met, only if the CuCrZr tube does not undergo softening due to overaging or irradiation creep at a maximum operation temperature. The upper temperature limit to avoid such effect lies between 300 °C–350 °C. The corresponding heat flux load would range from 14 to 18 MW m⁻². Should a part of the tube be subjected to still higher temperatures beyond this limit in long-term operation, the CuCrZr tube would possibly undergo a significant plastic fatigue, due to softening by irradiation creep and overaging.

Zoom In Zoom Out Reset image size

In contrast to the CuCrZr cooling tube, the soft copper interlayer was shown to undergo heavy plastic fatigue (i.e. low cycle fatigue) due to large variation of plastic deformation already existing in the early stage of HHF loading.

This feature is illustrated in figure 6, where the periodic temporal variation of three principal plastic strains (in a cylindrical coordinate system) is plotted. It is found that the radial and axial components exhibit quite large plastic strain amplitudes, whereas the hoop component has a roughly constant magnitude. The fatigue lifetime (the number of cycles to failure due to plastic fatigue) can be estimated using the plastic strain amplitudes on the basis of Manson–Coffin type fatigue life data. For the present ITER-type target model, the design estimates (predicted fatigue lifetime divided by 20) cannot fulfill the ITER load specification. This result indicates that the plastic fatigue of the Cu interlayer may pose a critical design concern in terms of the structural reliability of the whole target PFC.

Zoom In Zoom Out Reset image size

Several recent HHF test campaigns performed on ITER divertor target mock-ups demonstrated that W armor blocks were vulnerable to macroscopic cracking when the applied cyclic HHF loads were raised to 20 MW m⁻², which corresponds to the peak heat flux during slow transient events (loss of plasma detachment) [96]. HHF fatigue loads at 20 MW m⁻² often produced a deep crack through individual W blocks. The cracks were normally initiated on the top surface being heated, and grew further downward—often even reaching the cooling tube. The mechanism of the deep cracking phenomena was clarified by the authors in a computational study based on fracture mechanics, where the crack initiation was attributed to plastic fatigue, while the subsequent crack growth was shown to be a result of tensile stress concentration at the crack tip [97]. The strong tensile stress field at the crack tip appears during the cooling stage until the crack grows up to ca. 4.5 mm long depth downwards. On the other hand, if the crack becomes larger than 5.5 mm, then the tensile stress field at the crack tip develops during the HHF heating stage. This fracture feature is visually illustrated in figure 7 where the thermal stress fields (horizontal component) in the cross section of the target PFC are plotted for the HHF loading (20 MW m⁻²) and the cooling (150 °C) phase, respectively. The stress fields are presented for three different sizes of cracks (1.5, 3.5 and 5.5 mm). It is seen that the upper region of the armor is compressed during the HHF loading (crack is closed) while tensile stress states prevail in this region during the cooling (crack is opened). Figure 7 reveals that the stress state and the intensity of the singular stress field at the crack tip varies with the crack depth.

Zoom In Zoom Out Reset image size

Figure 8(a) shows the J-integral values at the crack tip computed as a function of crack lengths ranging from 0.5 mm to 5.5 mm from the top surface [97]. J-integral is a direct measure of crack tip loading, indicating the driving force for crack extension. In figure 8(a), the J-integral values at two positions of the crack front (free surface end and symmetry center) are presented for the HHF loading (20 MW m⁻²) and the cooling (150 °C) case, respectively.

Zoom In Zoom Out Reset image size

Figure 8(a) shows that the crack tip J-integral values during the cooling stage are sufficiently high compared to the critical fracture energy (or toughness) of the W armor (ca. 0.25 mJ mm⁻²) in the depth range up to ca. 4.5 mm so that the crack can grow further up to 4.5 mm in the repeated cooling phases, once it has been initiated at the surface. On the other hand, figure 8(b) shows that during the HHF heating phase the J-integral begins to increase rapidly only after the crack size has reached 5.5 mm reaching the critical values.

It is noted that the surface crack is initiated by plastic fatigue. In an extreme HHF loading case (20 MW m⁻²) as assumed above, significant fatigue may take place in the surface layer since this region is subjected to large plastic strain amplitudes and tensile stress in the initial transient stage of cooling. The fatigue life estimated for the HHF load of 20 MW m⁻² is less than 90 load cycles. Such a significant fatigue effect is attributed to the low yield stress of W in the surface layer caused by recrystallization.

W armor will also be exposed to characteristic thermal shock loads during transient instabilities such as edge localized modes (ELMs). The thermal loads caused by ELMs are predicted to be frequent (∼1 Hz), short (∼1 ms) and extremely intense (∼1 GW m⁻²), releasing huge amounts of thermal power (∼1 MJ) onto the surface. In many HHF fatigue tests simulating ELM-like thermal shocks, characteristic surface cracking features were observed [98, 99]. They showed that a network of many short cracks was created on the W armor surface where the typical crack depth ranged from several tens of μm to a few hundreds of μm.

In computational fracture mechanics studies, the crack patterns observed experimentally could be reproduced by use of J-integral calculation and XFEM simulation [100–102]. The XFEM study showed that the crack formed within the heat-loaded area is short and straight, whereas the crack near the boundary is long, deflecting its path from the vertical to horizontal orientation parallel to the surface. The origin of the driving force of surface cracking was the strong tensile stress field (radial component) which builds up as plastically induced residual stress. This tensile residual stress is a result of the compressive radial plastic strain (1.8%) produced locally within the heat-loaded area during ELM-like thermal shock loading. In contrast, the outer region surrounding the central heat-loaded area is deformed only elastically. When the HHF shock pulse is turned off, the thermal strain in the outer region relaxes elastically, trying to recover its original shape, while the plastic strain in the heat-loaded area remains as permanent deformation. Thus, the elastic recovery of the surrounding region exerts a tensile traction onto the heat-loaded surface area in the radial direction during cooling. The resulting tensile stress is higher than 1000 MPa, which is strong enough to cause crack initiation. This mechanism is illustrated in figure 9, where the profile of the radial component of the plastic strain and radial stress are plotted for three different time points, viz. at the end of the pulse (1 ms), shortly after the pulse shut-off (2 ms), and after cooling (10 s) [100, 101].

Zoom In Zoom Out Reset image size

First wall failure can be caused by thermal and/or mechanical stresses triggered e.g. by a plasma disruption, a temperature increase in the structure, or a pressurization of the coolant channels.

In case of a plasma disruption, eddy currents are induced, circulating through the mechanical structures and generating large electro-magnetic forces, as well as thermal loads (arising from ohmic heating), which in turn drive damage to the first wall.

In case of an ex-vessel LOCA, a double guillotine break in a large pipe is assumed leading to a coolant spill out (if no redundancy is foreseen) and air ingresses into the loop (see figure 11). In this context the grace time for the intervention of the fast plasma shutdown system (FPSS) has to be evaluated in such a manner that the integrity of the first wall structures is assured well before critical material conditions are reached [105]. In this context, the strength reduction by the structural temperature rise and melting point of the material are essential parameters describing the first wall integrity. If the integrity fails, in-vessel LOCA occurs as an aggravating failure. This multiple failure event is categorized as a design extension condition (DEC) according to [106].

In case of a LOFA caused by several means such as circulator seizure, malfunction of valves, clogging in first wall-breeding blanket (FW-BB) channels or piping or instrumentation & control failure, the first wall temperature and also pressure of cooling channels will rise without the FPSS. The pressure should be released to the pressure control system (PCS). Otherwise, if the pressurization exceeds the design limit, it will mechanically affect the FW-BB integrity considerably before reaching the critical temperature. If FW-BB integrity fails, the in-vessel LOCA also progresses as an aggravating failure, which subsequently triggers other failures. A failure of the PCS, spurious opening of relief valves, relief device internal leak or instrumentation & control erratic In/Out signal, LOHS and in-box LOCA (see figure 11) may also end up in an in-vessel LOCA, if the FPSS is not activated. Additionally, a large rupture and a leak of the sealing weld in the first wall are two representative PIEs for an in-vessel LOCA [105].

Figure 12 illustrates schematically the temporal evolution of an in-vessel LOCA. It occurs during normal operation at typical state variables as q₀, p₀, T₀ and full nuclear power. Considering a response time of pressure sensor for a pressure measurement, the LOCA can likely be detected at t_r (e.g. t_r   =  2.0 s in the HELOKA-HP facility [107]). The activation of the FPSS takes a time of ${{t}_{f}}-{{t}_{r}}$ (e.g. ${{t}_{f}}-{{t}_{r}}=3.0$ s in ITER) to terminate the plasma. In the shutdown state during a time t_f the decay heat still heats the structures, although it is less than  ∼2% of the nuclear power in normal operation. The FPSS itself causes plasma disruption producing a peak surface heat load of q_max within a disruption time of ${{t}_{d}}-{{t}_{f}}$ (e.g. ${{t}_{d}}-{{t}_{f}}=1.0$ s in ITER). The decay heat rapidly declines immediately after the plasma shut-down, but it continues to feed thermal power into the vacuum vessel for seconds, minutes, hours, days or even longer, depending on the material composition, which is calculated through a neutronic analysis. Depending on the first wall break size the coolant pressure drops and the coolant ingresses into the VV. If the pressure limit of the VV is reached, it will be discharged to an expansion volume (EV, for gas) or to a VV pressure suppression system (VVPSS, for liquid). If the LOCA is not detected and the plasma burns continuously further, the first wall temperature can increase up to the melting point, which is reached in the time at t_m . The mitigation with the FPSS stops the temperature rise and the first wall integrity can be assured.

Zoom In Zoom Out Reset image size

LOCA analyses have been conducted in several FPP concept studies since the safety and environmental assessment of fusion power have been recognized being a vital part for the public acceptance of fusion power. Herein, the SEAFP program in 1990 [103] can be considered as the first addressing the analysis of accident sequences with respect to safety aspects. From the analysis results an evaluation of accident consequence with respect to dose rate and its influence on the public can be conducted.

Since DEMO is still in pre-conceptual design phase, and safety analyses based on its four blanket concepts are on-going, up to now only accident sequence analyses are available, which correspond mainly to those already reported in the power plant conceptual study (PPCS) [108] and essentially covers mainly the blanket aspects, since the blanket covers more than 80% of the plasma confining area. Nevertheless, the PPCS study provides the most representative sequence evolutions, results and consequences, and can be still adopted as reference.

Within the PPCS study five different blanket model concepts have been investigated (see [108] and [109]). The breeding blanket concepts called model A (water cooled lead lithium—WCLL), B (helium cooled pebble bed—HCPB), AB (helium cooled lead lithium—HCLL) and C (dual coolant lead lithium—DCLL) are involved in the current DEMO development as well. Within the PPCS for the divertor in models B, AB and C helium has been considered as coolant, while in the current European DEMO approach water is favoured irrespective of the blanket concept. A comparison of all events studied for the five models in the view of the maximum attainable radiologic dose showed that a LOFA inducing subsequently an in-vessel LOCA in model B (HCPB) leads to the highest dose value among all LOCAs.

The postulated initiator of the event is an assumed pump trip in one of the FW-BB heat transport systems, leading to a loss of flow in one of the cooling loops, without a pump coast-down. Additionally, it is assumed that the FPSS does not intervene. As a consequence, the surface temperature of the PFCs increases continually until the critical temperature of the plasma-facing first wall component (EUROFER at 1073 K) is reached, and a break in the in-vessel FW-BB cooling channels happens. When the helium enters the VV, a plasma disruption occurs, transferring an energy of 6.0 MJm⁻² onto 418.0 m² of the first wall in 1 s. The FW-BB of the failed loop is first cooled down with residual coolant present before the complete drainage. The helium ingress into the VV causes a pressure rise up to a level of 1 bar; beyond this value the rupture disc fails and releases the gas towards the EV. The EV has a tentative free volume of $68\,000$ m³ (initial conditions of the air internal atmosphere are 30.0 °C and 0.09 MPa). During the entire leak period 2000 kg of helium (92% of the primary inventory) are discharged through the break into the whole containment. Finally, the overpressure in the VV should be safely mitigated under the design pressure of 0.2 MPa.

Leakages from the VV and EV building towards the external environment have to be considered. Radioactive materials inventories and their mobilization for model B are: tritium in the VV of 1 kg, dust of 10 kg with 7.6 kg stainless steel dust and 2.4 kg tungsten dust, and tritium in coolant of 1 g per loop.

Concerning the releases to the outside environment for this sequence, large amounts of the initial inventories are dispersed 24 h after the beginning of the accident. The major source is the high daily leakage (75%) of the EV. A reduction of the EV leakage rate from 75% of the volume per day down to 1% would be more effective than using a detritiation system (DS). As matter of fact, the adoption of the DS in addition to a decrease of the EV leakage rate to 1% vol./day reduces the tritium release further by 55%, and also the dust release by 30%.

It is important to assess the consequences to the public in terms of doses of other potential consequences. For the German regulation the dose criterion is defined as committed effective dose equivalent for the first seven days of exposure. For this LOFA accident, the dose value for the seven-day dose to MEI at 1000 m distance (24 h release, 95% fractile) was calculated to be 0.42 mSv. This dose value is far below the dose criterion for evacuation of 50 mSv, which is recommended by the International Commission on Radiological Protection (ICRP-60) and also used in many national regulations and for ITER. The maximum radiological dose to the public arising from the most severe conceivable accident driven by in-plant energies (bounding temperature sequence) was 18.1 mSv for the PPCS model B. It has to be stated that this value is obtained for a rather hypothetical event sequence assuming a total loss of cooling from all loops in the plant. Additionally, no active cooling, no operational functionality of active safety systems is anticipated, and further no human intervention whatever is considered for a prolonged period. The only assumed rejection of decay heat is by heat conduction and radiation through the layers and across the gaps of the model, towards the outer ambient, where likely a heat sink is provided by convective circulation of the building atmosphere.

Starting with Aveston et al [152], toughening in brittle matrix composites has been extensively studied for many years (see for example [143, 153, 154]). Having brittle tungsten as a matrix material, the behavior of W_f/W composites follow the same rules, and has been described based on this work [155–157].

Brittle materials do not possess the possibility for stress redistribution. This means if the local stress exceeds the local strength, e.g. at a crack tip, the whole system fails. Toughening by fibre-reinforcement works through the introduction of mechanisms for energy dissipation. This allows the relaxation of stress peaks, and thus increases the resistance against crack growth and therefore the toughness. As these mechanisms are not an intrinsic property of the material, but provided externally, they are also known as extrinsic toughening. If there is no possibility to improve the crack growing resistance of a material itself, this is the only way to improve the toughness of brittle materials [158, 159].

The contribution of these mechanisms $\Delta G$ to the toughness acts behind the crack tip, and thus lowers (by dissipation) the crack driving energy G, rather than directly influencing the matrix material at the crack tip. The following equation describes this relation:

$$\begin{eqnarray}G- \Delta G\left\{\begin{array}{*{35}{l}} <{{G}_{\text{c},\text{matrix}}} & \Rightarrow \text{crack}~\text{halts} \\ >{{G}_{\text{c},\text{matrix}}} & \Rightarrow \text{crack}~\text{propagates} \end{array}\right. \end{eqnarray} \tag{ 5 }$$

with
G	: energy release rate for crack driving
$\Delta G$	: consumed energy by extrinsic mechanisms
${{G}_{\text{c},\text{matrix}}}$	: critical energy release rate of matrix

Many different mechanisms have been reported, e.g. toughening by reinforcements like whiskers or fibres or microstructural mechanisms like transformation or matrix micro-cracking [143]. In this context, toughening by fibres is reported as the most effective mechanism (see table 2).

If a crack bypasses such fibres, they cause a traction force t on the crack faces, hindering further growth of the crack. The contribution to the toughening $\Delta {{G}_{f}}$ can therefore be determined according to equation (6) by integrating the contributions of all fibres ¹⁵ .

$$\begin{eqnarray} &&\Delta {{G}_{f}}={{V}_{f}}\cdot w_{f}^{\ast}={{V}_{f}}{\int}_{0}^{{{u}_{\text{max}}}}t(u)\text{d}u\end{eqnarray} \tag{ 6 }$$

with Vf	: volume fraction of fibres	(—)
$w_{f}^{\ast}$	: specific energy consumption	(J m−2)
umax	: maximum crack opening	(m)
t(u)	: traction of reinforcements on crack faces	(MPa)

An example of this behavior are ceramic fibre reinforced ceramics, which show a superior fracture toughness compared to the bulk ceramics, although their constituents are fully brittle [31, 144]. As a consequence, toughening is maintained even if the intrinsic properties of the material become degraded. In the case of W_f/W this was shown for model systems [155] and for bulk material [160]. There, the fracture toughness could be significantly improved by mechanisms not relying on any bulk material plasticity.

Fibre.

The tungsten fibres used in W_f/W are made of commercial drawn tungsten wire, which is characterized by high tensile strength, remarkable flexibility and notable ductility. This is attributed to a unique microstructure of elongated fine grains caused by the drawing process during fabrication. In figure 13, typical stress-strain curves obtained in tension tests of as-fabricated and annealed W wire doped with several 10 ppm potassium (K) are shown [161]. K doped W wires show ductile behavior up to an annealing temperature of 2173 K (30 min). The tensile strength stays up to about 2000 MPa for this temperature. At an annealing temperature of 2500 K (30 min) secondary grain growth leads to embrittlement. Such K doped wire has a long tradition in the illumination industry [162]. During the drawing process, K forms small bubbles which are aligned along the grain boundaries and pinning these, leading to superior temperature stability (more details in [163]). Tensile tests have also been used to investigate the mechanical properties of pure tungsten wire [164]. In the as-fabricated case and after annealing at 1273 K (3 h), pure W wire shows ductile behavior and a strength of 2900 MPa and $1900\pm 2$ MPa respectively. A heat treatment of 1900 K (30 min) leads to embrittlement, and a mean strength of approximately $900\pm 30$ MPa.

Zoom In Zoom Out Reset image size

Besides this, a key benefit is the ductility combined with a large local plasticity. The plastic deformation of the wire was identified as an important contribution to the toughening in the as-fabricated state of W_f/W [155]. A length of 250 μm was determined as region of increased plastic deformation in single fibre tension tests [165]. Furthermore, W fibres exhibit due to their metallic nature a rather high thermal conductivity (thermal conductivity of pure W: 173 W m⁻¹ K⁻¹ at 293 K, 147 W m⁻¹ K⁻¹ at 573 K [124]) in comparison with other high-performance fibrous reinforcements, such as SiC fibres. The toughness and flexural flexibility of W fibres furthermore enables rather easy handling, and allows that W fibres can be processed by textile technologies in order to produce fibrous preforms [156, 166]. All the above mentioned beneficial properties make W fibres, from the outset, a rather attractive fibrous reinforcement for W_f/W or composite materials in general.

Interface.

The interface between fibre and matrix in W_f/W composites has several functions. It has to provide a physical barrier between fibre and matrix, and thus maintains the composite structure during fabrication and operation. In addition, the interface must provide the load transfer between fibre and matrix and should in this sense be as strong as possible. On the other hand, it is essential that the interface, rather than the fibre, fails during crack propagation, to allow for the toughening mechanisms becoming active [167]. This phenomenon can be quantified by the so-called debonding criterion. For W_f/W where fibre and matrix feature the same elastic properties, the following equation is valid:

$$\begin{eqnarray}&&\alpha =\frac{\Delta {{G}_{i}}}{\Delta {{G}_{f}}}<0.25\end{eqnarray} \tag{ 7 }$$

with

$\Delta {{G}_{i}}$	: fracture toughness interface	(J m−2)
$\Delta {{G}_{f}}$	: fracture toughness fibre	(J m−2)

The interface system in W_f/W has been studied by means of push-out tests for various interfaces [148, 168, 169] and at different annealing stages [170]. As a first step, several interfaces were chosen according to interface types described for fibre-reinforced ceramics (see [171]). Fibre push-out tests were used on model systems consisting of a single fibre embedded into a tungsten matrix, so-called single fibre composites (see figure 14), to evaluate the applicability for W_f/W and to study the microscopic failure mechanisms. The interfaces were deposited on tungsten fibres using magnetron sputter deposition, and were afterwards coated with tungsten by means of chemical vapour deposition. Oxide coatings, i.e. $\text{Zr}{{\text{O}}_{2}}$ and $\text{E}{{\text{r}}_{2}}{{\text{O}}_{3}}$, with different thicknesses and multilayer coatings (with W or Zr), as well as pure Er, Cu and C were used.

Zoom In Zoom Out Reset image size

The interfacial parameters were evaluated by means of fitting the measured data to theoretical equations. This was combined with an extensive microscopic analysis. All interface systems were identified to fulfil the debonding criterion, since all determined values are smaller than $\alpha <0.05$. This is more than five times smaller than the theoretically optimal value and leaves therefore room for optimization.

Composite synthesis.

As shown above, the properties and functionality of the fibres and the interface system determine those of the W_f/W composite system. Therefore, it is important to maintain these during the synthesis process. For the synthesis, typically a fibrous preform is formed as a first step. This can be done by a winding [166] or a weaving process [156]. The interface system is either applied before this process or on the final preform. Afterwards the matrix is deposited on this preform to form W_f/W composites.

The very high melting point and high temperature strength of tungsten makes the application of classical composite production routes like liquid infiltration (more examples in [28]) difficult. The preferred manufacturing technique for tungsten is a powder metallurgical route (70–80% of total production) [13, 22]. During these processes usually high pressure and temperature are applied. The typical pressure range is 200–400 MPa and the typical temperature range is 2273–3273 K. Such conditions have a huge impact on the fibre properties (e.g. due to microstructural changes) and interface integrity (e.g. by thermal decomposition).

Jasper et al [172, 173] started investigations on the applicability of powder metallurgy for the production of W_f/W with respect to the effect on the fibres and the interface system. Hot isostatic pressing was used to produce model systems of single fibres embedded into a tungsten matrix. Besides the microstructural analysis they performed push-out tests to investigate the interface performance. First samples have been successfully produced at various temperature between 1900 K and 2200 K and a pressure of 200 MPa. It was shown that an erbia interface can withstand these conditions, but that its structure is strongly influenced. The pure tungsten wire used in the tests is fully recrystallised during the process.

An alternative processing route for tungsten is chemical deposition from the vapor phase, which is readily used for the production of coatings [174]. In this new composite production route technique tungsten hexafluoride ($\text{W}{{\text{F}}_{6}}$) is reduced by hydrogen (H₂) in a heterogeneous surface reaction, to produce solid tungsten. The process is conducted at a temperature between 573 K and 1073 K. In the case of ${{\text{W}}_{\text{f}}}/\text{W}$ typical temperatures are between 673 K and 873 K [148, 149]. The tungsten deposit is formed according to the following reaction:

$$\begin{eqnarray}&&\begin{array}{*{35}{l}} \text{W}{{\text{F}}_{6}}\left(\text{g}\right)+3\,{{\text{H}}_{2}}\left(\text{g}\right)\overset{673-873\,\text{K}}{{\longrightarrow}}\text{W}\left(\text{s}\right)+6\,\text{HF}\left(\text{g}\right) \end{array}.\end{eqnarray} \tag{ 8 }$$

The main advantages of this process are the comparably low temperature and the force-free character which allows preservation of the interface and fibre integrity, as well as fibre topology.

Surface deposition processes (chemical vapor deposition—CVD) have been used to produce cylindrical model systems containing a single fibre and with a final diameter up to 2.5 mm (see for example figure 14 or [170]). This easy-to-process model system was used to study fibre–matrix interaction, and has also been used in interface studies (compare section 7.2.2). Riesch et al [149, 166, 175] combined the chemical process with an infiltration step, and successfully fabricated bulk W_f/W for the first time. In a dual step process, gas flow and temperature are varied in such a way that material with very low residual porosity (<5%) is produced. The distance between the fibres is approximately 100 μm, leading to a fibre-volume fraction of about 0.3. In figure 15 a cross section of such a material is shown.

Zoom In Zoom Out Reset image size

Recently, a stepwise deposition of single tungsten wire layers has been established, to produce larger samples [156]. Here, the final sample is built of ten fibre planes, which are grown together step by step. The final sample has a volume of $50\times 50\times 3.5$ mm³ and a porosity of 4–7%. The fibre distance is 100 μm within the layers and about 150 μm between the layers, resulting in a fibre-volume fraction of about 0.2.

In summary, CVD processes have the advantage of low production temperature and the absence of mechanical impact, and are therefore ideal for W_f/W production. However, the experience in tungsten bulk production is low. For example, it is unknown whether it will be possible to influence the microstructure or the composition in a desired way. In contrast to this the production and processing techniques for powder metallurgical (PM) production routes are highly developed, as this is the standard process for tungsten bulk production. This opens up potential for easier optimization of the matrix properties, e.g. by combination of fibre reinforcement with the use of alloys (generally also brittle phases), e.g. self-passivating tungsten [177].

Mechanical tests have been conducted mainly in order to investigate the toughness and the active toughening mechanisms in W_f/W. Tension and bending tests on single fibre composites as well as on bulk samples have been performed. As-fabricated samples, as well as samples where the fibre has been embrittled by recrystallization and grain growth (this condition is called 'embrittled' in the following), have been tested.

The investigations of the single fibre composites have been combined with in situ synchrotron tomography to determine active toughening mechanisms qualitatively and quantitatively. By tension tests on single fibre composite samples, a high contribution of ductile fibre deformation to the toughness was proven in the as-fabricated case [155]. In bending tests, it was shown that the toughening also works in the embrittled condition, and the contribution of the individual mechanisms to the overall toughness has been determined using equation (6) [157]. For a fibre-volume fraction of 0.3 the ductile deformation of all fibres would lead to a fracture toughness above 150 MPa m^0.5. In the embrittled case, the main contribution to the increase in fracture toughness is by elastic bridging (up to 21 MPa m^0.5) and fibre pull-out (up to 36 MPa m^0.5). However, it has to be noted that contribution of the pull-out has been determined in a bending test, and might therefore be different in a real structure.

Bending tests on bulk samples have been used to determine the total increase of toughness. Tests have been conducted on as-fabricated samples produced by a CVI process [175], and on embrittled samples produced by a layered CVD process [160]. These tests revealed an intense toughening at room temperature and active toughening mechanisms in embrittled conditions. In both cases stable—i.e. controlled—crack propagation after crack initiation is observed. With ongoing crack propagation, the load bearing capacity of the material increases as well. This is typical for extrinsic toughening, as the mechanisms only become active behind the crack tip. In a conservative estimation based on the artificial crack length the fracture toughness was estimated to be ${{K}_{\text{Ic}}}=13.3$ MPa m^0.5 for the as-fabricated case and ${{K}_{\text{Ic}}}=9$ MPa m^0.5 for the embrittled case.

Recently Charpy impact tests have been conducted at room temperature [156]. An increased energy consumption is measured and almost all fibres fail ductile. Using linear elastic fracture mechanics, a fracture toughness of ${{K}_{\text{Ic}}}=140\pm 5$ MPa m^0.5 is calculated. From the results of the tests on model systems it is concluded that the main fraction of the toughness is gained by the energy consumption during the plastic deformation of the fibres.

In summary, the following toughening mechanisms have been observed in ${{\text{W}}_{\text{f}}}/\text{W}$ experimentally:

• debonding of interface between fibre and matrix;
• crack bridging by intact fibres;
• ductile deformation of fibres;
• fibre pull-out;
• crack deflection;
• crack meandering.

In addition, several effects are found within the matrix (e.g. debonding, pull-out of matrix elements) which are seen as an artefact of the imperfect fabrication process, and are aimed to be avoided in the future [149].

Crack bridging by intact fibres, the ductile deformation of fibres and fibre pull-out are described as the main mechanisms (compare figure 16). As long as the fibres are ductile (e.g. in the as-fabricated case) the toughening is dominated by the ductile deformation. If this ductility is lost, the main mechanisms are the elastic bridging and the pull-out of fibres. To visualize the potential of the new tungsten composite material the determined toughness values are shown in an Ashby diagram of toughness over Young's modulus (figure 17). The high fracture toughness values are mainly reached in the as-fabricated case, where the reported toughening is higher compared to the embrittled case.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For both face centred cubic (fcc) and bcc metals, cold rolling leads to an increase in strength combined with a decrease in ductility. This is the well-known state-of-the-art knowledge of materials. However, the behavior of W (and maybe also of molybdenum) does not fit into this classical picture. It was shown by Wei and Kecskes that cold rolling increases strength and ductility [182]. In their paper, they show that the lower the rolling temperature, the higher the strength and ductility. This behavior is a puzzling phenomenon, and needs further explanation.

The second property that is improved by cold rolling is toughness. Pippan et al showed that technically pure W foil with a thickness of 160 m in its as-received condition has a room temperature (RT) fracture toughness of 70 MPa m^0.5 (L-T direction) and 60 MPa m^0.5 (T-L direction), respectively [183, 184] (L-T and T-L means in-plane orientation, details of the code can be found in [185]). These values are very high compared to the well-known RT fracture toughness of W (5 MPa m^0.5) and are in the same regime as W 26 wt.% rhenium (Re) (60 MPa m^0.5, [54, 186]). Whereas for W 26 wt.% Re the origin of the high fracture toughness can be found in the formation of a beneficial solid solution, the high toughness of W foils can only be related to its microstructure, and the change of the microstructure during cold rolling. Finally, cold working has a significant impact on the BDT. Again W does not fit the classical picture, as for W the BDT can be shifted to a lower temperature through cold rolling [187] while, for example, the BDT of α-Fe increases after cold working.

In summary, the material response of W to cold rolling improves ductility, toughness and BDT, and therefore, highly deformed cold-rolled W foils appear to be interesting candidates for the synthesis of W laminates.

Discussing the extraordinary properties of W foil, size effects have to be considered [188–190]. These include, for example, microstructural size effects. The high degree of cold rolling leads to pancake-shaped grains with typical dimensions of $0.5\times 3\times 15$ μm³. This means that in the direction of the thickness of the foil, the grain size is in the so-called intermediate grain size (GS) regime; the ultra-fine grained (UFG) regime, which is between nano crystalline (GS  <  100 nm) and coarse grained (GS  >  1 m) [189]. Valiev et al showed that UFG materials produced by severe plastic deformation show paradoxically increased strength and ductility [191]. The underlining microstructural mechanisms of the increase in ductility are not yet understood, but are intensively discussed [192]. Considering microstructural size effects, one has to admit that even the most popular size effect, the Hall–Petch effect, cannot be completely explained in detail [193].

In fully recrystallized W foil another size effect comes into play. In this condition, the grain size equals or even exceeds the foil thickness. In other words, single grains span the whole cross-section from one surface to the other. This, in turn, enables another deformation mechanism. Here, dislocations can easily move (since there are no grain boundary obstacles) to the surface, where they are annihilated (figure 19, left). This mechanism increases the plasticity significantly and might also take place in W–Cu laminate interfaces [180, 181].

Zoom In Zoom Out Reset image size

Finally, especially for a W–Cu laminate, a geometrical size effect can be provoked, that is, if the Cu interlayer is smaller than 1 μm (figure 19, middle). According to the W–Cu phase diagram, there is (in good approximation) no diffusion of W into Cu and vice versa. So a W–Cu interface can be regarded as a sharp bcc–fcc interface. If the dislocations in the Cu interlayer are meant to glide, they are restricted on both sides by W layers. This forces the dislocations to move in the W channel, which might lead to a significant increase in strength of the Cu interlayer, and thus of the W–Cu laminate.

In the development of advanced ceramic materials, the toughness was successfully increased by applying ceramic fibres in a brittle ceramic matrix. The increase in the toughness can be explained by some bridging effects as well as by the energy dissipation through the creation of new surfaces along the crack paths. Energy dissipation by surface creation also takes place in W laminates, especially when samples are tested in the X-S orientation (stress field perpendicular to the interfaces). For such samples, the crack is subsequently deflected, and has to be initiated layer by layer [181]. This crack deflection not only comes along with the creation of a surface but also with a subsequent change in the stress state at the crack tip. A deflected crack moving away from the ligament plane results in a preferable stress tensor in terms of crack propagation.

However, the contribution of energy dissipation by crack deflection to the total energy dissipation should be negligible. Most energy is dissipated by plastic deformation of the foil—and here mainly of the W foil [86]. Especially when samples are oriented in the L-T direction (figure 19, right), crack deflection does not take place. So, an improvement in dissipated energy compared to the pure plate material must be sought in the mechanical properties of the W foil or the interlayer. Considering Pippan et al's results for the high fracture toughness of W foil at RT, it is no surprise that L-T oriented samples also show advanced fracture behavior.

It can be concluded that the fracture behavior of W laminates is complex and depends on several parameters, but that the excellent properties of cold rolled W foil might be the dominating aspect in the discussion of energy dissipation.

Radiation induced lattice defects impede glide motion of dislocations, resulting in increase in yield stress (radiation hardening), and also may change the structure and chemical compositions of random GBs from the controlled conditions, thereby causing radiation embrittlement. In order to suppress radiation embrittlement, it is necessary to introduce a high density of sinks for radiation induced defects. Effective sinks are GBs and dispersoids. Therefore it follows that a structure containing ultrafine grains and a high density of dispersoids is the target structure and called nanostructure.

TiC is one of the dispersoids effective for the above purpose; TiC precipitated in the grain interior will work as sinks for radiation-induced defects. A high density of GBs is introduced due to grain refinement by heavy plastic working at temperatures where recrystallization does not occur. However, dispersoids in metals disturb plastic working (dispersion strengthening), and grain refinement becomes more difficult with increase in dispersoids and decrease in grain size (grain size strengthening). Moreover, a high free energy due to fine-grained structures may lead to recrystallization and grain growth at elevated temperatures. It is therefore our main objective to achieve grain refinement with thermal stability in the recrystallized state, without relying on plastic working for dense composites.

In view of the target structure, the results on Mo–1.5 mol% TiC bicrystals can be an encouraging reference: the precipitation of TiC and segregation of the TiC constituents occur at each of all the GBs and grain interior. As the grain size decreases, the GB area significantly increases with 1/d (d being the grain diameter). Intergranular embrittlement will occur even when only one unreinforced GB is left, and hence all of the GBs must be reinforced by the behavior of TiC (segregation and precipitation).

Nanostructures are practically producible by the powder metallurgical (PM) route. In order to attain the target nanostructure by the PM route, the following microstructure control is required; (1) to produce alloyed powder of a solid solution of W–Ti–C with nanosized grains from the starting powders of W and TiC, and (2) to produce fully densified compacts of equiaxed ultra-fine grains of W matrix containing a high density of fine precipitates of TiC at GBs and grain interior from the alloyed powders of W–Ti–C.

If the oxygen level is appreciably high, the oxygen preferentially reacts with Ti to form titanium oxides because the standard free energy for the formation of titanium oxides is much lower than that of TiC [199]. As a result, carbon atoms left behind in the W matrix would react with the surrounding W atoms to form W₂ C (W₂ C is known to be a brittle phase): the nucleation and growth of a metal carbide or metal oxide requires diffusion of metal elements as well as carbon or oxygen. Since carbon or oxygen atoms are surrounded by the metal matrix atoms, of which diffusivity is much lower, it is difficult to attain the equilibrium state where the brittle ${{\text{W}}_{2}}\text{C}$ phase is not formed.

Mechanical alloying (MA). 

In order to attain the above item (1) MA will be the most effective: MA is the method for producing alloyed powder with nanosized grains (less than 30 nm) in solid solution of several elements that are thermodynamically insolvable in each other (or extremely limited in solubility) [200, 201]. Therefore, if the processed powder is not alloyed powder, but just comminuted and/or dispersed powder from the starting powder, the process should be called MD (Mechanical Dispersion), not MA.

Prior to the MA operation, the mixed powders are degassed by heating at around 1273 K for 1 h in vacuum. The degassed powders are put in a metal vessel with hard balls in a purified H₂ or Ar atmosphere and then subjected to high energy ball milling by rotating or vibrating the vessel in an ambient temperature. Figure 24 shows a schematic illustration of the MA process consisting of three stages of powder evolution, i.e. severe plastic deformation and work hardening, fracture and cold joining at freshly revealed fracture surfaces, and homogenization. Powder that has completed all these stages is called mechanically alloyed powder (MA powder).

Zoom In Zoom Out Reset image size

MA always admits contamination by metal impurities introduced from the milling vessels and balls used for MA. Therefore, the vessel and balls must be made of materials acceptable from the viewpoint of ductility, thermal stability, radioactivity, etc. The MA vessels and balls that the authors employed were made of TZM (Mo–0.5Ti–0.1Zr–0.01C).

It is also important to use high purity powders without any lubricants, and prevent contamination with gaseous interstitial impurities of oxygen and nitrogen contained in the atmospheres through the fabrication process. For this, the powders are always treated in a well degassed glove box filled with a purified Ar or H₂ gas.

Hot isostatic pressing (HIP).

MA-HIPed refractory alloys.

Since the low temperature ductility of MA-HIPed refractory alloys is considerably affected by the atmosphere in MA, we designate the compacts as Alloy/Ar or Alloy/H, which denotes the sealed gas in the MA process, Ar or H₂, respectively [42].

As stated in section 9.2.1, our main objective is to achieve grain refinement with thermal stability in the recrystallized state without relying on plastic working. On the other hand, for many years the authors had believed that ductilization in any PM fabricated nanostructured alloys requires plastic working (PW) such as hot forging and hot/warm rolling following MA and HIP. As a matter of fact, a nanostructured Mo–0.2% TiC/Ar alloy that was at first fabricated by MA-HIP exhibited an enhanced low temperature ductility following PW [203]. However, increased TiC addition up to 1.0 wt.% (1.5 mol% TiC same as figure 23) often disturbed PW and the benefit of PW was poor [204, 205].

Controlled neutron irradiations on a pre-embrittled Mo alloy imparted a very interesting and encouraging result [205–208]. When the MA-HIPed Mo–1.0% TiC/Ar with insufficient PW was exposed to controlled temperature-cycle irradiations in JMTR (Japan Materials Testing Reactor) between 573 and 773 K with fluence of 1 $\times {{10}^{24}}$ n m⁻² (E_n   >  1 MeV) and subjected to three-point impact bending (5 m s⁻¹) and Vickers microhardness tests, it exhibited remarkable ductilization regardless of significant radiation hardening (figure 25) [205–208]. This ductility enhancement is quite opposite to radiation embrittlement and was designated as RIDU (Radiation Induced DUctilization) [206, 207]. The only possible explanation for RIDU is that all of the possible cracking sites such as random GBs and interfaces of dispersoids are significantly reinforced by the neutron irradiation, and the beneficial effect of the reinforcement exceeds the embrittling effect of radiation hardening. The reinforcement due to neutron irradiation can be attributed to radiation-enhanced or -induced precipitation of Mo₂C, (Mo, Ti)₂C, TiC_x, (Ti, Mo)C_x and their segregation, which may occur preferentially at random GBs with higher energies [207, 208]. It is noted that radiation induced point defects distribute very homogeneously in the material, and significantly promote lattice diffusion towards the state of lower free energies. Similar results on RIDU were reported for Mo–0.2TiC/Ar, although the degree of RIDU is much smaller than that for Mo–1.0TiC/Ar [206, 209].

Zoom In Zoom Out Reset image size

The above reinforcement mechanism in RIDU is essentially the same as the one already shown in the unirradiated Mo bicrystals. Therefore, RIDU tells us that we have achieved the target nanostructure where all of the GBs are reinforced by the behavior of TiC and its solid solutions (segregation and precipitation).

Successful ductilization of nanostructured Mo alloys without PW and neutron irradiation after consolidation by HIP or spark plasma sintering (SPS) was performed by Takida et al [210]. They examined the effects of addition of 0.8 mol% transition metal carbides (TMCs) other than TiC on the DBTT through three-point impact bending. They found that the ductilization of Mo–0.8 mol% TMC/Ar depends strongly on the kind of dispersoids, and TaC addition appeared to be more effective than TiC addition. In this case, however, more addition of 1.6 mol% TMC resulted in less ductilization.

Efforts to fabricate nanostructured W alloys by MA-HIP-PW were also made. The capsule initially employed for HIP at higher temperatures was pure Ta with much higher melting point (3263 K) than mild steel (≈1760 K). However, PW for MA-HIPed W alloys was much more difficult than for MA-HIPed Mo alloys: Only the (0.2–0.3)% TiC added W alloys were subjected to sufficient PW, which lead to the relative density of 99.5% and an enhanced ductility at lower temperatures measured by three-point impact bending [211, 212].

The above results on enhanced ductility suggested that the ductilization of nanostructured refractory alloys can be achieved by increasing the relative density to  ≈99.5% or more. In order to increase the relative density of MA-HIPed W with TiC additions, therefore, an improvement in all the processes of MA and HIP was made to seek thorough removal of impurities that are likely responsible for less densification: especially important were degassing of the starting powder and MA powder and outgassing of the residual gas in the capsule (mild steel) charged with the MA powder prior to TIG welding of the capsule [202], as mentioned in paragraph HIP in Section 9.2.2.

As a result, HIPing at 1623–1673 K, approximately 2/5 of the melting point, resulted in essentially fully densified compacts with relative densities of 98–99% depending on the sealed gas in MA. The MA-HIPed compacts of W–(0.25–1.5)% TiC exhibited equiaxed grains with the average diameter of 50–200 nm in the recrystallized state and were designated as UFGR (Ultra-Fine Grained, Recrystallized) W–(0.25–1.5)TiC (/H and /Ar) [42, 202].

The UFGR W–(0.25–1.5)TiC has attracted great attention because such a nanostructure was the first to fabricate in W materials. Characterization of the UFGR W–(0.25–1.5)TiC was performed by domestic and international research collaborations. Regarding the ductilization, however, no appreciable ductility occurred at room temperature and the DBTT defined as the null-ductility temperature was as high as  ≈830 K. Figure 26 shows the effects of TiC addition, MA atmosphere and HIP pressure on three-point bend fracture stress at room temperature [42, 213]. TiC addition of 0.25% significantly increases the fracture stress, but further TiC additions up to 1.5% do not change the fracture stress. Moreover, HIPing for W–1.1TiC/Ar at the increased pressure of 1 GPa (UH) increases the fracture stress to 2.6 GPa mainly due to the significant suppression of nano-sized Ar bubbles. These results indicate that the beneficial effects of TiC additions are masked probably by residual pores and insufficient reinforcement of random GBs and dispersoid interfaces.

Zoom In Zoom Out Reset image size

Ductilization of materials does not occur until the fracture stress exceeds the yield stress. A rough estimate of the yield stress assuming the Hall Petch relation for W gives as high as 5–6 GPa for UFGR W with the grain size of  ≈100 nm. It seems quite difficult to enhance the fracture stress to above 5–6 GPa. Therefore, in order to achieve ductilization of nanostructured W, the following microstructure control is required in addition to MA-HIP. That is, to significantly promote the precipitation of TiC and segregation of the TiC constituents at all of the random GBs, grow the grains to appropriate sizes to reduce the extremely high yield stress due to grain size strengthening to the required level and thoroughly remove the residual pores. For this purpose, a new process for microstructural control, designated as grain boundary sliding-based microstructural modification (GSMM), was developed.

Grain boundary sliding-based microstructural modification (GSMM).

GSMM is based on activation of GB sliding without crack formation and its linkage, where the recrystallized, equiaxed grain geometry is maintained and the deformation is driven by active grain rotation and extensive relative displacement of the adjacent grains accompanied with moderate grain growth [42]. These GB activities would lead to the sufficient precipitation of TiC and segregation of the TiC constituents at the random GBs, thereby reinforcing the GBs with holding the recrystallized state. It may also have the effect of removing the residual pores through GB diffusion during GB sliding. In addition, the moderate grain growth caused by GSMM leads to the decrease in yield stress to an appropriate level.

GSMM is performed under the condition that the above GB activities are the largest. Such GB activities are expected to occur during superplastic deformation where sufficiently large elongation without crack formation and its linkage and relatively low flow resistance are assured. Therefore, the superplastic behavior of W–(0.25–1.5)TiC (/H and /Ar) was examined as a function of TiC content, MA atmosphere, temperature and strain rate (figure 27) [42, 214–216]. The optimum conditions of GSMM turned out to be (1.1  ∼  1.2)% TiC addition and plastic strain rate of 10⁻⁵ s⁻¹ at 1923–1973 K.

Zoom In Zoom Out Reset image size

Mechanical properties and microstructure.

GSMM processes by compression were applied to MA-HIPed (UFGR) W–1.1TiC (/H and /Ar) as a function of temperature and reduction ratio. It was found that significant increase in fracture stress and ductilization at room temperature are achieved when GSMM was performed at temperatures of 1923–1973 K and reduction ratio above 80% (figure 28) [42, 213]. The GSMM processed W–1.1TiC (/H and /Ar) is designated as TFGR (toughened, fine-grained, recrystallized) W–1.1TiC [42].

Zoom In Zoom Out Reset image size

The DBTT (nil ductility temperature) for TFGR W–1.1TiC/H is shown in figure 29 as a function of impurity contents of oxygen and nitrogen [217]. The DBTT decreases with decreasing oxygen content and is between 240 and 420 K for oxygen contents of 160–870 wppm. Since the DBTT of UFGR W–1.1TiC/H is  ≈830 K, GSMM leads to significant reduction in DBTT.

Zoom In Zoom Out Reset image size

Figure 30 shows that promoting segregation and precipitation of TiC, GSMM significantly increases the amount of dispersoids [42]. On the other hand, increase in oxygen content leads to the formation of the brittle W₂C phase, which may act as initiation sites of cracking. Since an excessive amount of oxygen disturbs the recombination of decomposed TiC through the formation of the W₂C phase and a Ti oxide as mentioned in section 9.2.1, it is necessary to control the oxygen content to an acceptably low level.

Zoom In Zoom Out Reset image size

TEM microstructures and GB orientations analyzed by electron back scatter diffraction (EBSD) patterns for the above UFGR and TFGR W–1.1TiC (/H and /Ar) are shown in figure 31. Leaving the grain shape equiaxed, GSMM at 1923 K increases the grain size almost by one order: 0.6 μm for W–1.1TiC/Ar and 1.5 μm for W–1.1TiC/H [42, 218, 219]. The distribution of most of the GB orientation obeys the Mackenzie curve based on the random distribution of grain orientation, indicating that most of the GBs are random GBs of high energy. No visible nano-sized bubbles are observed in TFGR W–1.1TiC (/H and /Ar).

Zoom In Zoom Out Reset image size

Effects of GSMM on the formation and growth of dispersoids in TFGR W–1.1TiC/H are shown in figure 32 [220]. The average size of dispersoids is larger at GBs than in grain interior, and increases with increasing GSMM temperature: 90 nm in the grain interior and 160 nm at the GBs for GSMM at 1923 K; 160 nm in the grain interior and 390 nm at the GBs for GSMM at 2273 K. In those GSMM processed specimens, the K-S orientation relationship was recognized at the interface between the TiC phase and W matrix [221].

Zoom In Zoom Out Reset image size

The main features of TFGR W–1.1TiC/H are listed in table 3, in comparison with commercially available pure W (e.g. ITER grade W).

Table 3. Features of TFGR (toughened, fine-grained, recrystallized) W–1.1TiC.

Characterization item	TFGR W–1.1TiC	Commercially available pure W
Microstructure	Equiaxed, fine grains	Elongated, relatively coarse grains
	TiC fine dispersoids	No dispersoids
	Recrystallized state	Stress-relieved state
	High sink density	Low sink density
Grain boundary in the recrystallized state	Significantly reinforced by TiC precipitation and its segregation	Weak (not reinforced)
Recrystallization temperature	Already recrystallized	1200–1300 °C (Tmelt 3410 °C)
DBTT in the recrystallized state	Around RT	higher than RT
Response to high heat loading	Resistant to cracking and surface roughening	Prone to cracking and surface roughening
Radiation damage	Resistant	Vulnerable

The densification (consolidation) of the MA powder can also be performed by the GSMM process. That is, the HIP process may be replaced with GSMM when consolidation by the GSMM is initially performed at a temperature similar to the HIP temperature (first step of GSMM), and GSMM is then performed as in the above MA-HIP-GSMM (second step of GSMM). The first and second steps of GSMM can be successively performed by merely changing the temperature. This more time saving and economical method, without HIP, is designated as the MA-GSMM process [217]. TEM microstructures and bend ductility at room temperature showed that MA-GSMM processed W–1.2% TiC/H exhibits essentially the same features as the MA-HIP-GSMM processed W alloys [217]: The grain structures are essentially recrystallized, equiaxed with average grain size of  ≈2 μm and TiC exists at GBs and grain interior with the K-S orientation relationship between the W matrix. Three tested specimens exhibit appreciable, reproducible ductility at room temperature.

Mechanical and plasma-wall interaction behavior.

The structural applications of the materials to the first wall require assured resistance to mechanical loads without any deformation and cracking. The performance of TFGR W–1.1TiC observed upon thermal shock loading and thermal fatigue loading, which are anticipated mainly in divertor environments, indicates its applicability in first wall environments, because thermal shock and thermal fatigue loadings dynamically and statically generate thermal internal stresses due to a temperature gradient in the material.

Thermal shock tests were conducted for TFGR W–1.1TiC/H with oxygen contents of 160 and 850 wppm and UFGR W–0.5TiC/H (oxygen content: 170 wppm) by using JUDITH-1 under the condition of ITER-ELM (edge-localized mode) like loading; t (pulse length)  =  1 ms, P (heat density)  =  1.1 GW m⁻² ($\Delta T$ (resultant surface temperature increase)  ≈2270 K), ${{T}_{\text{base}}}$ (base temperature)  =  373 K, n (repeated pulse number)  =  100. TFGR W–1.1TiC/H with 160 ppm oxygen exhibited no cracks on the surface, whereas TFGR W–1.1TiC/H with 850 ppm oxygen showed tiny and very shallow cracks on the surface, and UFGR W–0.5TiC/H exhibited net-pattern like cracks on the surface and a semi-circle flaw in the interior with a maximum depth of 0.17 mm [31]. TFGR W–1.1TiC/H with 160 ppm oxygen, even when the base temperature was decreased to room temperature, did not exhibit any cracks on the surface.

Thermal fatigue tests were performed for TFGR W–1.1TiC/H and ITER grade, stress relieved pure tungsten (W/SR) by using electron beam irradiation with the following two heating patterns imposed consecutively; (1) surface temperature of around 1973 K, duration of 180 s and (2) repeated irradiations of 2 s irradiation and 8 s rest in one cycle of 10 s for about 1 h in total, which resulted in temperature variations between 1523 and 723 K with 380 cycles. W/SR exhibited significant surface roughening (plastic deformation) and cracking, whereas TFGR W–1.1TiC/H did not exhibit any surface roughening or cracking (figure 33) [222]. The observed behavior of TFGR W–1.1TiC/H can be attributed to high dynamic yield stresses due to grain size strengthening and high fracture stresses due to reinforcement in GBs enriched with TiC and its constituents. These results suggest that TFGR W–1.1TiC/H will be resistant to mechanical loading to be generated in the first plasma wall environments.

Zoom In Zoom Out Reset image size

Behavior under neutron irradiation.

As far as the dispersoids and GBs are stable against neutron irradiation and effectively act as sinks for radiation induced point defects, the nanostructured W materials are expected to exhibit less hardening and less embrittlement due to neutron irradiation. A good example of the beneficial effects of nanostructures on the significant suppression of radiation embrittlement is found in oxide dispersion strengthened (ODS) ferritic steels irradiated with fast neutrons to $\left(0.3\sim 3.8\right)\times {{10}^{26}}$ nm⁻² at 646–845 K in Joyo operating at JAEA. The DBTTs measured by Charpy impact testing did not shift to higher temperatures by the irradiation [224].

Similar results on radiation hardening were found and are shown in figure 34 [208, 223] where the size and number density of radiation induced defects and the increase in Vickers microhardness by irradiation are compared between UFGR W–0.5TiC (/H and /Ar) and commercially available pure W in the stress relieved state (W/SR) irradiated at 873 K up to 2  ×  10²⁴ nm⁻² (E_n   >  1 MeV) in JMTR. The degrees of the microstructural changes and radiation hardening for the UFGR W–0.5TiC (/H and /Ar) were significantly reduced compared with those for the W/SR. These results reflect the beneficial effect of a high density of sinks contained in UFGR W–0.5TiC (/H and /Ar).

Zoom In Zoom Out Reset image size

Applicability of MA-HIP-GSMM to other nanostructured alloys.

We further examined the applicability of GSMM to ductilization in other nanostructured alloys that contain a high density of random GBs and dispersoids and a small amount of residual pores, but do not exhibit intergranular embrittlement like group VIA transition metals. Two alloys, V–1.4Y–8W–0.8TiC and SUS316L–2TiC, were fabricated by MA-HIP followed by either simple annealing in vacuum or GSMM process, and then tensile tested at room temperature at an initial strain rate of 1  ×  10⁻³ s⁻¹. The results of tensile tests are listed in table 4. The uniform and total elongations are significantly increased by GSMM in comparison with those by simple annealing. This ductility improvement is most likely due to the reinforcement of the interfaces between dispersoids and the metal matrix. Therefore, we can say that the MA-(HIP)-GSMM is not only the key process for ductilization in nanostructured W materials, but can also be widely applied to many other nanostructured alloys that do not exhibit intergranular embrittlement like group VIA transition metals.

Table 4. Effects of GSMM (grain boundary sliding-based microstructure modification) on tensile ductility at room temperature in other nanostrucured alloys that do not exhibit intergranular embrittlment like group VIA transition metals.

Material	Thermo /mechanical treatment	Yield stress, ${{\sigma}_{0.2}}$ (GPa)	Tensile stress (GPa)	Uniform elongation (%)	Total elongation (%)
V–1.4Y–8W–0.8TiC	MA-HIP followed by anneal at 1300 °C	0.71	0.72	5	9
V–1.4Y–8W–0.8TiC	MA-HIP followed by GSMM at 1300 °C	0.66	0.70	14	20
SUS316L–2TiC	MA-HIP followed by anneal at 950 °C	0.68	0.95	10	10
SUS316L–2TiC	MA-HIP followed by GSMM at 950 °C	0.85	1.13	20	23

The hydrogen isotope (protium, deuterium and tritium) permeation through a single-layer material can be expressed in principle by a multiplication of two predominant processes: solution and diffusion [259]. When hydrogen atoms form a solid solution in the material, the hydrogen solubility S (mol m⁻³) is given by Sieverts' law:

$$\begin{eqnarray}&&S={{K}_{\text{S}\,}}{{p}^{0.5}}\end{eqnarray} \tag{ 10 }$$

where K_S is the equilibrium solution constant (reciprocal number of Sieverts' constant, mol m⁻³ Pa^−0.5) and p is the hydrogen pressure (Pa). The hydrogen diffusivity D (m² s⁻¹) is a thermally activated process expressed by the Arrhenius rate equation:

$$\begin{eqnarray}&&D={{D}_{0}}\exp (-\frac{{{E}_{\text{D}}}}{RT})\end{eqnarray} \tag{ 11 }$$

where E_D (J mol⁻¹) is the activation energy of diffusion, R (8.314 J K⁻¹ mol⁻¹) is the gas constant, and T (K) is the temperature. The hydrogen permeation flux per unit area at steady-state J (mol m⁻² s⁻¹) can be described by the first Fick's law:

$$\begin{eqnarray}&&J=D\frac{\partial c}{\partial x}\end{eqnarray} \tag{ 12 }$$

where c (mol m⁻³) is the hydrogen concentration at position x (m). Given a sufficiently lower concentration at the outlet side than that at the inlet side of the material, J through the material of d in thickness is expressed with the product of the equations (10) and (12):

$$\begin{eqnarray}&&J={{K}_{\text{S}}}\,D\,\frac{{{p}^{0.5}}}{d}\end{eqnarray} \tag{ 13 }$$

where ${{K}_{\text{S}}}D$ is the permeability P (mol m⁻¹ s⁻¹ Pa^−0.5) as the intrinsic parameter of the permeation. P is also a thermally activated process expressed as:

$$\begin{eqnarray}&&P={{P}_{0}}\exp (-\frac{{{E}_{\text{P}}}}{RT})\end{eqnarray} \tag{ 14 }$$

where E_P (J mol⁻¹) is the activation energy of permeation calculated by the sum of E_D and the enthalpy difference between hydrogen in gas phase and that in the material.

It should be noted that the simplest case of the permeation, where the rate-limiting process is the diffusion of hydrogen atoms through the solid, can satisfy the equations described above. When the permeation rate is limited by a hydrogen molecule process as typified by surface reactions such as absorption, dissociation, recombination and desorption, the exponent of the driving pressure p will increase to unity (n  =  1). The surface contribution would be derived from a large surface roughness, chemical reactions with the material or impurities such as carbon [285]. When the diffusion and surface limited regimes are comparable, the exponent of the driving pressure will indicate between 0.5 and 1.

The permeation barrier performance of a coated sample is often evaluated using a permeation reduction factor (PRF): a ratio of P or J of an uncoated sample to that of a coated one. However, it is also important to pay attention to absolute values of the permeability and permeation flux, since the PRF varies depending on not only the barrier coating but the substrate material: e.g. austenitic and ferritic steels.

For the development of a TPB for fusion reactors, the clarification and control of dominant factors influencing hydrogen permeation under reactor operational conditions are essential. The most basic factor would be types of chemical bonds derived from different electronic structures. Metals, except for a few elements such as tungsten and gold, generally have higher permeability than ceramics because metallic bonds exert much lower solubility and much higher diffusivity than ionic and covalent bonds. That is why ceramics have been employed as TPB candidate materials. However, the number of studies focused on the discussion of hydrogen permeation behaviours through ceramics is quite small in comparison with metals. Here we compile investigations on hydrogen permeation mechanism in ceramic bulks and thin films.

The major factors influencing hydrogen permeation through ceramic materials may include composition, crystal structure, phase and defect. As mentioned previously, oxides and carbides are representative compositions for TPB coatings. Therefore, the dominant parameter influenced by composition is hydrogen's interaction with oxygen or carbon as well as metal or non-metal elements. In fact, Yao et al reported that C–H and O–H bonds made an important contribution to permeation barrier efficiency from the analysis using secondary ion mass spectroscopy (SIMS) and infrared spectroscopy (IR) for TiC and $\text{Si}{{\text{O}}_{2}}$ coatings [288]. Regarding the influence of the crystal phase, tritium diffusivity and deuterium solubility of several types of α-SiC and β-SiC was investigated by Causey et al [289]. The results indicate that the phase makes smaller effects in permeation than other factors such as purity and crystallinity. Chikada et al performed deuterium permeation experiments for amorphous SiC coatings deposited on steels, indicating the amorphous structure also exhibited an efficient permeation reduction [285]. That is probably because Si and C chemically bind to deuterium, and the bindings block the migration of other deuterium atoms. In the case of $\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}$, Levchuk et al showed deuterium permeation through amorphous and γ-$\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}$ mixture coatings and an α-$\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}$ coating [282]. The mixture coating showed higher permeability than the α-$\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}$ one due to a higher hydrogen diffusivity in the amorphous structure than that in the crystallized structure. These results clearly indicate that both chemical composition and crystallinity strongly influence hydrogen permeation in the coatings.

The defect in a crystal lattice might be one of the most important factors in permeation through permeation barriers. In particular, grain boundaries, which are a kind of aggregation of point defects, have a clear influence on permeation behaviours. Zakharov et al showed that polycrystalline W and Mo tubes had somewhat higher permeabilities than those of single crystal in the temperature range of 400–1200 °C [290]. In the case of ceramics, Chikada et al investigated deuterium permeation through $\text{E}{{\text{r}}_{2}}{{\text{O}}_{3}}$ coatings fabricated by filtered vacuum arc deposition [291]. The results of permeation experiments and microstructure observation indicated that the deuterium permeation was dominated by grain boundary diffusion in the temperature range of 500–700 °C. This suggestion was verified by deuterium concentration in the coatings measured using nuclear reaction analysis (NRA) and nano-scale SIMS [292]. On the other hand, the hydrogen permeation through sintered $\text{A}{{\text{l}}_{2}}{{\text{O}}_{3}}$ tubes tested at higher temperatures (1200–1450 °C) showed that the dominant transport process was lattice diffusion [293, 294]. Therefore, the dominant transport process, lattice diffusion or grain boundary diffusion, is dependent on the temperature range and the grain size. One criterion to determine the dominant process is to estimate the activation energy of permeation E_P in equation (14), since in principle the grain boundary diffusion has lower values of E_P.

Two kinds of actively water cooled W/Cu FGM based HHFC mock ups have been designed, borrowing ideas from the W–Cu joint concepts, viz. the monoblock concept and flat tile concept, which were selected by ITER. For W–Cu joint mock-ups, it has been demonstrated that the monoblock design has better performance than the flat tile design under high heat flux loads. The W/Cu FGM monoblock was designed for examining and comparing the temperature distribution and thermal stress under high heat flux to that of W/Cu FGM flat tile mock-up. A geometric analysis of the models is shown in figure 50. Model 1 is a design of monoblock concept (figure 50(a)). The cooling channel is located in the middle of the W/Cu FGM. Distance from bottom of cooling channel to bottom of mock-up is 3 mm. Model 2 is a flat tile concept design (figure 50(b)). The HHFC is formed by three sections of different materials: at the top, the W armor faces the high heat load and radiation; at the bottom, the Cu substrate removes heat by means of water flowing through the cooling channel; in the middle, the graded layers between the W armor and the Cu substrate mitigate the thermal stress. The volume fraction of W in the graded layers is expressed by the following equation:

$$\begin{eqnarray}&&{{V}_{\text{W}}}=1-{{(x/h)}^{p}}\end{eqnarray} \tag{ 15 }$$

in which V_W represents the volume fraction of the W phase in the FGM, p is known as the gradient index, h is the thickness of the FGM, x is the distance to the pure W surface. Here x  =  0 corresponds to pure W and x  =  h to pure Cu. The different gradient architectures used to join the Cu and W and the thermo-mechanical properties of Cu, W and FGM are listed in tables of [319] and [320].

Zoom In Zoom Out Reset image size

The operation condition is modeled and analyzed according to the design of ITER. A steady-state heat load of 5 MW m⁻² is supposed to load on the top of the W armor. Water is flowing through the channel in the Cu at 160 °C, 4 MPa and 10 ms⁻¹. Stress and strain fields are obtained from the resulting temperature profile.

The temperature gradient and the von Mises stress gradient distribution of W/Cu FGM based HHFC with the monoblock type and the flat tile type have been demonstrated. The maximum temperature occurs on the W surface exposed to the heat and decreases towards the channel. For flat tile type, the maximum temperature is more controlled by number and thickness of graded layers. Increasing the thickness of graded layers from 1 to 3 mm raises the maximum temperature from 558 K to 594 K. The maximum temperature for the monoblock type is 593.7 K, nearly the same as for the flat tile type. The stresses in the interlayers tend to be compressive under operational conditions. The stress is more uniformly distributed throughout the graded interlayer of monoblock type, as compared with the sharper stress gradient in the flat type. Clearly, the optimum stress distribution for the operation stresses in flat type exists for the three layers in thickness of 3 mm. It also shows the smallest plastic strain. The von Mises stress, ${{\sigma}_{22}}$, and equivalent plastic strain of monoblock type and flat type are nearly the same [319].

The potential of functionally graded W/EUROFER97 joints has been examined using the finite element method in several publications [46, 321]. Here it shall be illustrated for the helium-cooled divertor design developed at KIT [322]. Elasto-plastic and elasto-viscoplastic simulations [323] have been performed under variation of the layer thickness, layer orientation and transition function. The resulting stresses and strains were used as basis for the subsequent lifetime estimation. Cycling of temperature comes along with the design of a tokamak reactor, and yields failure of the joint due to fatigue and creep. For the calculation of the number of allowable cycles, fatigue as well as creep damage are considered in the evaluation for the EUROFER97 part. Creep is neglected for the tungsten part. The lifetime of the joining layer could not be assessed due to the lack of experimental data. However, the quantified improvement in the lifetimes of the joined parts when varying the parameters of the joints also applies for the joining layer, at least qualitatively. The materials parameters, boundary conditions, governing equations and details used in the FEM model can be found in [46] and [323]. The model geometry is a tube with an inner diameter of 14 mm and an outer diameter of 15 mm.

A cooldown from 800 °C to 0 °C and subsequent thermal cycling between 0 °C and 650 °C for 100 cycles results in the maximum equivalent inelastic strain shown in figure 51. While very high deformations can be expected in a concentrated region at the outer side of the joint for thin interlayer thicknesses, the strain is far less, and distributed over a bigger area for thick graded interlayers.

Zoom In Zoom Out Reset image size

The result of the fatigue-lifetime estimation for the EUROFER97 region is shown in figure 52. The results for the tungsten region can be found in [323]. The Basquin–Manson–Coffin relationship between the equivalent strain range $\Delta \epsilon$ and the number of allowed cycles N_d are proposed as design criteria in [324]. Thereby $\Delta \epsilon$ is calculated according to the RCC-MR code. While a brazed joint only allows a cycling over several hundred cycles according to these design criteria, several ten thousand cycles are allowed when a linear joint transition is considered.

Zoom In Zoom Out Reset image size

The calculated lifetime considering creep is presented in figure 53. It contains a curve for EUROFER97 based on the design criteria proposed in [324], and two curves for EUROFER97 and ODS-EUROFER, respectively, based on the time to rupture curves [325]. A detailed description of the parameter extrapolation for ODS-EUROFER97 can be found in [323]. The lifetime is calculated through a linear damage accumulation for each time step of the FEM simulation. Thereby the integration point with the shortest lifetime is considered.

Zoom In Zoom Out Reset image size

Looking at the number of cycles allowed, a clear positive correlation to the interlayer thickness is again noticeable. Nevertheless, a saturation level is reached at about 1 mm layer thickness. At higher thicknesses the maximum stresses in the EUROFER97 section are not any longer at the interface. Instead, bending stresses caused by the radius difference become important. Only much thicker layers (>5 mm) would further reduce the stresses. A second insight is that the oxide dispersed strengthened version of the EUROFER97 steel does not substantially increase the lifetime. The explanation for this phenomenon is that the creep of EUROFER97 is strain-controlled, rather than stress-controlled.

The FE simulations performed show that a functionally graded joint between tungsten and EUROFER97 can drastically decrease the thermal mismatch stresses and strains occurring in the divertor component considered, and thus improve its failure behavior during thermal cycling. However, the functionally graded layer shall have a sufficient thickness that has to be taken into account in the selection of fabrication methods.

Smoothing the gradation on the EUROFER97 side or beveling of the linear graded layer reduces the maximum equivalent inelastic strain up to a factor of two (shown in [46]). However, since the realization of both is fairly challenging this gain might not justify the development efforts required. Simulations on W/EUROFER97 FGM are also performed for its application as erosion protective coating for First Wall component using the finite element code ABAQUS. The residual stresses induced in the whole component during the fabrication phase and subsequent operation phase are calculated for determining thicknesses of the W/EUROFER97 FG-layer.

In the simulations of the first cooling down phase, the behaviour of all materials is considered to isotropic, linear elastic and perfectly plastic. For the simulations of the operation phase, the Norton power law of creep is taken into account. The 2D sketch mesh, generalized plane strain elements, as well as boundary condition are introduced in detail in Ref. [326]. The model is loaded by a homogeneous temperature field that varies over time, simulating a cooling down after the hot manufacturing process. After this phase the temperature alternates between 20 °C and 600 °C with a dwell time of 24 h at 600 °C, which corresponds to an operation cycle of the fusion reactor.

As shown in figure 54(a), taking the model with a 0.5 mm thick FG-layer as an example, the maximum von Mises stress that occurs in W, which is limited by the yield stress, gradually reduces from W to EUROFER97. In addition, the stress in W is compressive, while tensile stresses appear in EUROFER97. The compressive stresses in W are not significantly destructive, even though they are almost equal to the yield stress of W. This means that W is supposed to deform plastically during the cooling phase. In contrast, the tensile stresses in EUROFER97 are lower than its yield stresses. Plotting the maximum of von Mises stress and equivalent plastic strain over the thickness of FG-layer—see figure 54(b)—the maximum equivalent plastic strain in W obviously decreases with increasing thickness of the FG-layer, while that in EUROFER97 is equal to 0 for all modelled thicknesses of the FG-layer. The maximum von Mises stress in EUROFER97 is limited by the yield stress for 4 mm thick FG-layer, which means that the thickness should be less than 4 mm to avoid plastic deformation of EUROFER97.

Zoom In Zoom Out Reset image size

Figure 55(a) illustrates the variation of the equivalent creep strain after 365 cycles (one year) with a hold time of 365  ×  24 h over all FG-layer thicknesses. The equivalent creep strain decreases drastically with increasing thickness of FG-layer. It is worth mentioning that the equivalent creep strain is less than 10⁻⁵ after 365 thermal cycles when the thickness of FG-layer is thicker than 1.2 mm, and it is almost the same for the model with a 1.5 mm thick FG-layer. As it can be seen in figure 55(b), the maximum equivalent creep strain of all investigated thicknesses increases over thermal cycles, but the thicker the FG-layer, the smaller the creep strain. Particularly for FG-layers thicker than 0.7 mm, a shakedown is observed. In addition, the maximum equivalent creep strain becomes small and constant over the number of cycles for FG-layer thicker than 1.0 mm. The maximum equivalent creep strain is less than 10⁻⁵ even after 600 thermal cycles when the thickness of FG-layer is 1.2 mm. The allowable number of cycles in EUROFER97 seems to increase proportionally for 1.2 and 1.5 mm thick FG-layers after creep damage accumulation of 600 cycles. As shown in figure 56, the allowable number of cycles N_cr increases linearly to the thicknesses of the FG-layer when it is thinner than 1.0 mm, while it increases similarly over one order of magnitude for the models with 1.2 and 1.5 mm thick FG-layers. The lifetimes improve remarkably with increasing thickness of the FG-layer.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

A novel one step method, called 'resistance sintering under ultra-high pressure (RSUHP)', has been developed to fabricate W/Cu FGM without the addition of any sintering additive. W and Cu powders were mixed and milled with different compositions and then stacked layer by layer in a steel mold to form a green compact. The W/Cu FGM green compact was placed in the pressure vessel; the details of the setup can be found in [308]. First, the pressure was mechanically cubic isostatic loaded on the W/Cu FGM green compact, then an alternating current (AC) was applied to the sample, thus heating and sintering the FGM by Joule heating. Sintering parameters should be suitable; investigations on sintering parameters and sintering mechanism are introduced in [309, 327, 328].

Figure 57(a) shows the picture of the fabricated W/Cu FGM mock up by RSUHP. Figure 57(b) illustrates the backscattering image of the cross section of the overall five-layered W/Cu FGM with p  =  1, in which the W component is white and the Cu is black colored. A good graded composition transition is found. The interfaces between layers are clear. This reveals that there has been no obvious composition migration during the short sintering time. The relative density of pure W layer was more than 97%. The hardness of the pure W layer was 780 HV, which is much higher than pure conventional sintered W (about 350 HV). This should be due to the finer grain size of the produced W than that of the conventional sintered W. The thermal conductivity of this W/Cu FGM was 113 W m⁻¹ K⁻¹ at room temperature. In addition, six-layered W/Cu FGM, W/Cu/W FGM, and Cu/W/Cu FGM could be fabricated by RSUHP [309]. This means that FGMs with any designed combination could be obtained by means of RSUHP.

Zoom In Zoom Out Reset image size

A Nd:YAG laser beam with a 10 ms pulse length, wavelength of 1064 nm and pulse frequency of 5 Hz was used for high heat load test on W/Cu FGM fabricated by RSUHP and infiltration welding. The laser beam diameter was about 2 mm. Argon was used to protect the samples from oxidation during the test.

After the high heat flux tests, micro cracking but no cratering occurred on the surface of the tungsten at the condition with laser energy density of 198 MW m⁻², as shown in figure 58(a). It should be noted that obvious cratering occurred on pure tungsten samples fabricated by RSUHP at higher energy density. This demonstrates that W/Cu FGM has higher thermal resistance than pure tungsten. With higher energy density of 260 MW m⁻² and duration of 10 s, crater morphology was found in the laser dot for all samples. The crater is enlarged when the hot impact time is increased, as shown in figure 58(b), revealing larger scale evaporation during the impact process [327, 328].

Zoom In Zoom Out Reset image size

Tungsten powders with average particle sizes of 3, 7, 15 μm and a purity of  >99.5%, and Cu powder with a particle size less than 45 μm, and a purity of more than 99.9% were used as the starting materials. The processing includes three steps: (1) Fabrication a porous graded W skeleton; (2) Infiltration of the W skeleton with liquid Cu to form a W–Cu graded composite; (3) Joining W plate on the W rich surface of the W–Cu graded composite by using the rapidly solidified ribbon-type Ti based amorphous filler with a melting temperature of 850 °C [319]. The details of the steps have been recorded in [327].

It is found that the compositions have graded distribution in W–Cu graded composite, as shown in figure 59(a). Figure 59(b) shows the interface between pure tungsten and W–Cu graded composite joined by brazing. It can be seen that both interfaces between filler-metal with pure tungsten and W–Cu graded composite are homogeneous and pore-free. This demonstrates a very good wettability of the molten brazing alloy on the tungsten. The thermal conductivity of this W/Cu FGM was 155 W m⁻¹ K⁻¹ at room temperature. After thermal shock tests introduced in section 12.2.1, no obvious damage occurred, but heavy oxidation occurred in the graded layer after 30 cycles of quenching. No change occurred to the surface of the tungsten with W/Cu FGM after infiltration welding for 2 min hot impact in air with laser energy density of 198 MW m⁻² introduced in section 12.2.1.

Zoom In Zoom Out Reset image size

In terms of FGM heat sink applications for a water-cooled divertor target, W/CuCrZr composite metals produced by liquid melt infiltration have been characterized extensively [329]. Figure 60(a) illustrates the thermal conductivity of the W/CuCrZr composite metals with three different compositions, namely 30 vol.%, 50 vol.%, and 70 vol.% W. It is seen that the composite with a W content of 30 vol.% exhibits a quite high thermal conductivity of about 300 W m⁻¹ K⁻¹ over a wide temperature range up to 600 °C. With an increasing W content, the thermal conductivity decreases accordingly but still yields notably higher conductivity values compared to pure W (thermal conductivity of pure W: 173 W m⁻¹ K⁻¹ at 20 °C, 147 W m⁻¹ K⁻¹ at 300 °C [11]).

Zoom In Zoom Out Reset image size

Figure 60(b) shows typical tensile stress-strain curves of a W/CuCrZr composite metal with 50 vol.% W at three test temperatures [329]. The curves indicate a considerable strengthening effect compared to the corresponding tensile strength values of CuCrZr alloy used as matrix material. The ultimate tensile stress values are notably higher compared to the monolithic CuCrZr alloy (minimum ultimate tensile stress of CuCrZr: $\leqslant 452$ MPa at 20 °C, $\leqslant 325$ MPa at 300 °C [11]). However, it can be seen at the same time that the increase in strength does not come cost-free, as the ductility is notably reduced compared to monolithic CuCrZr (average total elongation of CuCrZr: $\geqslant 20.3$% at 20 °C, $\geqslant 10.5$ at 300 °C [11]). Nevertheless, the latter feature does not seem to be a critical issue, as these ultimate tensile strain values are still substantially higher than the maximum strain that is expected to occur in the heat sink of a highly loaded PFC. For instance, in the case of an ITER-like divertor target under DEMO-relevant loading conditions, the peak plastic strain is predicted to be less than 1% according to finite element analysis (FEA) [95].

In order to verify the expected thermal and mechanical performance, as well as the structural reliability of the W/CuCrZr composite metals as FGM heat sink material, HHF fatigue tests were performed [330]. To this end, water-cooled demonstrator mock-ups with a rather simplified design were manufactured, as illustrated in figure 61(a) [330].

Zoom In Zoom Out Reset image size

They consisted of castellated flat tungsten tiles, a CuCrZr heat sink block, and a functionally graded W/CuCrZr interlayer bonded between the W armor and the CuCrZr heat sink. The W/CuCrZr composite interlayer consisted of three layers, differing in composition, with a thickness of approximately 1.5 mm for each layer. Details about the manufacturing process of the composite FGM, as well as of the mock-ups, can be found in [329, 330]. During the HHF fatigue tests, cyclic heat flux loads were applied to each of the three mock-ups at the hydrogen beam irradiation facility GLADIS at IPP Garching [138]. All three mock-ups withstood HHF loads of 15 MW m⁻², while one of the mock-ups could be tested even up to 20 MW m⁻², verifying the thermal and structural performance of the FGM heat sink as well as the joint mock-up successfully. Figure 62 shows a micrograph of the FGM interlayer after the HHF test illustrating fully intact microstructure without any visible cracks or defects. This result delivers the proof of principle demonstrating the high technological potential of melt infiltrated W/Cu composite metals as FGM heat sink materials for the design of a highly loaded PFC [330].

Zoom In Zoom Out Reset image size

Several initial sizes of W powder were investigated for plasma spraying, and the results show that the smaller the size of initial powder, the lower porosity of the coatings will be. The interlayer coating materials of W/Cu were obtained by mixing W and Cu powder with volume ratio of 50:50 and linearly graded [317, 331]. W/Cu mixtures and W powders were sprayed layer by layer on an oxygen-free copper substrate by atmosphere plasma spraying (APS) using a PT-A-3000 S plasma-spray facility made in Sweden. Argon was used as the plasma gas. Argon and nitrogen were used for cooling the substrates and preventing the coatings from oxidation. The detailed spraying parameters are given in [331]. The thickness of the W/Cu graded layer was 300 μm, and the maximum thickness of pure W coating was 3 mm.

As shown in figure 63, the W/Cu interlayer coating exhibits the typical lamellar structure, formed by overlapping W and Cu splats, with the tendency to form a layered structure. The porosity of the graded interlayer was about 5%, and the porosity of the pure W layer was about 12%. The bonding strength between the substrate and the coating was 23 MPa. The thermal conductivity of this W/Cu FGM was 57 W m⁻¹ K⁻¹ at room temperature. A Nd:YAG laser beam was used for high heat loading test on these specimens. Under the laser energy density of 152 MW m⁻², almost no change occurred on all the coatings. But when the laser energy density was increased to 198 MW m⁻², cracking, melting, and cratering appeared in all coatings after 240 cycles. Under the same conditions after 1200 cycles, peeling occurred on the surface of one sample and no peeling occurred on the coating surface of the other samples [331].

Zoom In Zoom Out Reset image size

An electron beam with a 2 ms pulse width was used for high heat load tests. The energy density was 10 MW m⁻². After 1000 cycles of hot impact, microcracks occurred in the W coating of the W/Cu FGM by APS and delamination occurred between the coating and the substrate. No obvious cracks occurred in the W/Cu FGM fabricated by the other two methods (RSUHP and infiltration-welding).

The large difference of the melting point (about 1900 °C) between tungsten and steel impedes conventional sintering, due to the lack of an overlap in sintering temperatures. Nevertheless, the successful sintering of both metals has been demonstrated by RSUHP and optimized in [339].

W powder with an average particle size of 2 μm and EUROFER97 powder (spheroidized in nitrogen) with a particle size of d₅₀  =  50 μm served as raw materials. Both powders are mixed in an agate mortar with different volume ratios according to the composition design. In this work, samples with four layers at different compositions, ranging from 20 to 80 vol.% EUROFER97 in 20 vol.% steps, were fabricated, see figure 64(a). The avoidance of too high temperatures or even the melting of EUROFER97 particles is of major importance for the selection of process parameters. Thus the temperature was chosen to be less than 1200 °C, which is much lower than the normal sintering temperature of W powders in the absence of any sintering aid. Nevertheless, the temperature at the interface of W particles should be very high, due to the high contact resistance on the particle interface; thus surface diffusion or even surface melting may occur quickly, and particles are fused together in the presence of strong current. The thickness of the outer two layers was 3 mm, and for the inner two layers it was 0.5 mm. The microscopic analysis revealed a microstructure of good quality, free of cracks and very low porosity (<2%).

Zoom In Zoom Out Reset image size

A drawback is the current role of matrix (tungsten) and particulate (EUROFER97), since the composite behaves almost as tungsten. The opposite case, a composite whose ductile steel matrix hosts hard tungsten particles could improve the ductility of this composite.

In this experiment [312] a graded layer could be realized by a four step multilayer design (figure 64(b)). It should be noted that the realization of a continuous step-less gradation is also feasible. The microstructure is amorphous or nanocrystalline (depending on the mixing ratio). This very important aspect offers the chance to profit from the superior materials properties due to the ultra-fine grained/nanocrystalline microstructure. All sputtered coatings exhibit a dense microstructure with no remarkable porosity after a slight polishing with oxide polishing suspension. However, a major drawback of magnetron sputtering is the thickness of sputtered tungsten coatings, which is limited to  ∼30 μm [340].

In early experiments, graded W/steel composites on W substrates have already been successfully produced by VPS [341]. Unfortunately, a major drawback was the high porosity. A high material density very close to the theoretical density and a high thermal conductivity are essential when used as coating for a plasma facing component. An improved density was achieved in [312], being well below 4 vol.%. The multilayer structure of VPS composites, typical for coatings in an early development stage, could also be achieved by an appropriate focussing of both particle trajectories. Nevertheless, the adhesion of these VPS coatings on the substrate was insufficient. The reason for the poor interfacial strength might be that the tungsten substrate roughness was not high enough, as bonding is only formed by mechanical clamping. A systematic roughening or sculpturing of the substrate surface can improve the adhesion [309]. To overcome the lack of interfacial strength, a thin vanadium foil with a thickness of 20 μm was used in [312] for the bonding to WL10. This does not result in an additional production step, as the vanadium interlayer can be joined to the tungsten using the same diffusion bonding parameters as for the EUROFER97 side. Such an interlayer is necessary, since diffusion at 800 °C between two tungsten surfaces is not sufficient.

Functionally graded coatings were produced [312] with a thickness of up to 1 mm (figure 64(c)), but thicker tungsten-based coatings up to several millimeters are also possible. Based on the images taken by optical and scanning electron microscopy, a microstructure of good quality with no cracks and a porosity  <4% can be observed. However, cavities with a size up to 20 μm are strewn along the sample volume.

Heat treatments at different temperatures for one hour revealed insights on the thermal stability of the sprayed tungsten/EUROFER97 composites. For all the tested temperatures up to 1100 °C, no cracks were visible by the SEM, either in the EUROFER97 or in the tungsten regions [312]. However, gaps between both materials arose when the temperature was equal to or higher than 1000 °C for one hour.

VPS is suitable for the production of functionally graded EUROFER97/tungsten coatings, which can act as interlayers for joining EUROFER97 and tungsten bulk material by diffusion bonding at 800 °C. No phase formation or change in chemical composition in the functionally graded layers could be detected after the bonding process. The produced joints were thermally cycled ten times between 20 °C and 650 °C, and proved their basic capability to withstand the operating conditions in a future fusion reactor. This is an improvement in comparison to the direct diffusion bonding of EUROFER97 and tungsten at 1050 °C, where all samples fell apart in the post bonding heat treatment [342]. Certainly, the low welding temperature of 800 °C, which avoids the austenite phase formation of EUROFER97 at T  >  800 °C, and reduces thermally induced stresses during the cooling down, is a very positive circumstance. In addition, the gradation helps to reduce thermally induced macroscopic stresses as it has been assessed numerically in [46].

The application of functionally graded tungsten/steel material is not limited to divertor components, since it is also considered as erosion protective coating for the plasma-facing first wall.

Based on the simulation results summarized in section 12.1.2, FG tungsten/EUROFER97 coating systems on EUROFER97 substrates are fabricated by VPS for FW application. Three and five layers with linear variation of the compositional ratio are introduced as representatives of stepwise linear gradation of FG-layer. The as-received surface morphology and cross-sectional structures of both types are shown in figure 65. The surfaces have a smooth structure, which indicates that particles melted well during the spraying process and re-solidified, forming a pancake-like structure. In addition, several spherical particles of about 10 μm exist on the surface, which are un-melted tungsten particles. The cross-sectional SEM micrographs show that stepwise linear gradation is formed successfully between the tungsten coating and the EUROFER97 substrate, and that there is no obvious interface between layers of the FG-layer, especially for coating with five layers, which has a finer gradation. Nominal and true concentrations of tungsten in percent by volume are investigated by using EDS, as shown in figure 66(a). True concentrations of tungsten are quite consistent for the case of 25% and 63% nominal value, while the true values are a little lower than the nominal value for other cases. However, the true concentrations as desired follow a (globally) linear gradation over the number of layers as it can be seen in figure 66(a). The thicknesses of the coatings with three layers are shown in figure 66(b). No delamination is observed for both coatings [343].

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As shown in figure 67, coatings are characterized by re-solidified pancake-like deposition with columnar structure of tungsten. The size of the columnar crystals is 2–4 μm in height, and the columnar crystal is radially oriented perpendicular to the substrate. In addition, un-melted particles and irregular pores of several micrometers in size exist in the tungsten coating, as well as in the FG-layer. Most pores occur around un-melted particles and at interfaces among lamellar layers. It is noted that no cracks or delamination could be observed in tungsten coating and FG-layer. It is assumed that the porous structure constrains the propagation of fissures or cracks [341].

Zoom In Zoom Out Reset image size

The optimization of spraying parameters and vacuum accompanied by proper melting and re-solidification of powders yields the lowest porosity of all coatings. In particular, the porosity of the tungsten coating is  ∼4% according to the image analysis, and that of the FG-layer is less than 2%. It seems that the low melting point and ductility of EUROFER97 powders provide flowing of melted metal, which yields low porosity. Despite some gap-like defects, sound interfaces between the FG-layer and the EUROFER97 substrate are observed (see figure 67) [343].